Does timing matter? The role of health information shocks in measuring willingness to pay

Brinkmann, Carolin; Neumann-Böhme, Sebastian; Brouwer, Werner B. F.; Stargardt, Tom Article — Published Version Does timing matter? The role of health information shocks in measuring willingness to pay The European Journal of Health Economics Provided in Cooperation with: Springer Nature Suggested Citation: Brinkmann, Carolin; Neumann-Böhme, Sebastian; Brouwer, Werner B. F.; Stargardt, Tom (2025) : Does timing matter? The role of health information shocks in measuring willingness to pay, The European Journal of Health Economics, ISSN 1618-7601, Springer, Berlin, Heidelberg, Vol. 26, Iss. 8, pp. 1401-1413, https://doi.org/10.1007/s10198-025-01774-7 This Version is available at: https://hdl.handle.net/10419/330818 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ ORIGINAL PAPER The European Journal of Health Economics (2025) 26:1401–1413 https://doi.org/10.1007/s10198-025-01774-7 willingness-to-pay (WTP) can help policymakers assess value for money and prioritize different interventions based on individual preferences as part of their allocative decisions [1–5]. In healthcare, this approach is especially important because the absence of market prices and the desire of policymakers to align their decisions with the preferences of citizens mean that value is usually not observable. WTP therefore has become an established measure in preventive and curative health interventions [6, 7]. Yet despite their broad use and theoretical underpinnings, the validity of WTP estimates has been criticised, for example due to their insensitivity to scale [8] and hypothetical bias [9]. Additionally, WTP values are known to be associated with the elicitation procedure, framing, and study design [10–18]. Introduction To quantify the value of non-marketed goods and services such as health care interventions, health economic evaluations often seek to elicit the maximum amount that individuals are willing to pay for them. This amount can then be used to estimate the value of the interventions in monetary terms, facilitating comparisons between them. In this way, Carolin Brinkmann [email protected] 1 Hamburg Center for Health Economics, University of Hamburg, Hamburg, Germany 2 Erasmus School of Health Policy and Management, Erasmus University Rotterdam, Rotterdam, The Netherlands Abstract Objectives The optimal point in time to measure willingness-to-pay (WTP) remains unclear. We investigated the role of health information shocks (HIS) in individuals’ WTP, analyzing the extent to which news of SARS-CoV-2 infections among people they know/themselves altered WTP for booster vaccinations. Methods We elicited WTP in eight European countries using the European Covid Survey. First, we presented participants with a hypothetical setting recommending a booster vaccination that had to be paid out-of-pocket. To measure WTP, we elicited a lower and upper WTP limit, and a WTP value contingent on both of these. To measure HIS, we asked about the duration since participants received news of COVID-19 cases among people they know (including themselves), as well as the degree of personal connection to these cases and their severity. We used a two-part model to estimate the association between HIS and individuals’ WTP. Results Among the 5809 observations, 76.8% stated a WTP for a booster vaccination greater than €0. At least one HIS was reported by 61.9% of participants. The occurrence of a HIS was associated with an increase in WTP of €14.54 (logistic: P <.0001, gamma: P =.1493) compared to no HIS. The WTP was higher when the HIS occurred in the four weeks before the survey. Controlling for socio-demographic and COVID-19 covariates decreased significance and effect sizes. Conclusion Our findings suggest that a recent HIS is associated with a higher probability of having a positive WTP. Timing, in relation to some relevant event, therefore may matter when measuring WTP for health interventions. If so, finding the optimal point in time to measure WTP is difficult and may depend on the policy question under consideration. Keywords Willingness to pay · Information shock · Health shock · Vaccination JEL code I10 · I12 Received: 22 April 2024 / Accepted: 24 March 2025 / Published online: 17 April 2025 © The Author(s) 2025 Does timing matter? The role of health information shocks in measuring willingness to pay CarolinBrinkmann1· SebastianNeumann-Böhme1,2 · Werner B. F.Brouwer2· TomStargardt1 1 3 C. Brinkmann et al. Another issue to consider when designing WTP exercises in the context of health interventions is when to ask participants for a valuation. Just as the phenomenon that adaptation may affect subjective health-related quality of life [19–22], the timing of a WTP exercise in relation to the occurrence of some relevant ‘shock’ may be associated with its results. This may especially be the case if participants have experienced a health event or health information shock. The concept of information shocks has been the subject of recent health research [23–27]. In the present study, we aimed to contribute to this literature by exploring the role of health information shocks (HIS) in relation to WTP. To do so, we focused on the example of SARS-CoV-2 booster vaccinations, eliciting the hypothetical WTP for these in eight European countries and investigating the extent to which WTP was associated with the timing of a HIS. We defined a HIS either as having experienced a COVID-19 infection oneself or having heard of infections among people one knows. At the time of data collection in December 2021 and January 2022, total cumulative COVID-19 cases were between 8,180.7 (Germany) and 22,777.1 per 100,000 persons (United Kingdom) in Europe [28], i.e., a majority of the population had not experienced a COVID-19 infection (yet). While the Delta variant was predominant in incident cases in late December 2021, the more contagious omicron variant only started to spread in Europe in early 2022 [29]. Our results suggest that the occurrence of a HIS is associated with WTP estimates, especially in the short run. A HIS might prompt individuals to be willing to pay for the offered good. The observed association suggests that the timing of WTP surveys might influence their results and ultimately affect allocation decisions. Methods The European Covid Survey The European Covid Survey (ECOS) was an online survey that elicited information on the attitudes, behavior, worries, and health characteristics of European residents aged 18 years or older approximately every two months between April 2020 and December 2022. The samples consisted of approximately 1000 participants per country and wave and were representative of the general population in each country in terms of gender and age. The participating countries were Denmark, France, Germany, Italy, Portugal, the Netherlands, the United Kingdom (UK), and, after July 2021, Spain. Participants were recruited by the market research company Dynata. Participation was anonymous and required written informed consent. The questionnaires were developed in English, translated into the various national languages, and piloted in 10% of each sample. Further information can be found elsewhere [30–33]. The present analysis is based on the 9th wave of ECOS data, collected between 23 December 2021 and 11 January 2022. Measuring health information shocks (HIS) We defined health information shocks (HIS) as important new information received by participants about their own health or the health of someone in their social environment. Based on this definition, we measured HIS by asking participants about (a) the duration since they heard news of COVID-19 cases among people they know (including themselves), (b) their degree of personal connection to these cases, and (c) the severity of these cases. To do so, we used a matrix in which participants could tick boxes for each combination of temporal (when? ) and relational (who?) information (see Online Resource). The duration since receiving news about a case of COVID-19 was shown using two-week intervals ranging from “0 to 2 weeks” to “7 to 8 weeks”, and complemented by an open interval of “more than 8 weeks”. The degree of personal connection to these cases was shown as “household”, “family”, “friends”, “colleagues”, and “no” or “don’t know”. Self-shocks (infections among participants themselves) were implicitly included in the category “household”. Lastly, participants were asked to indicate the overall severity of the majority of the COVID-19 cases, with the options ranging from “more severe than expected” to “milder than expected”, and including a mixed category (“mixed, i.e., half of the cases more severe than expected, and half of the cases milder than expected”). When analyzing the data, we treated the category “Don’t know” the same as the category “No” based on the assumption that a case of COVID-19 that the participant did not remember or did not consciously perceive would be the same as that of the absence of a case of COVID-19. WTP measurement To measure WTP, we presented participants with a hypothetical scenario in which (a) a booster vaccination was recommended, (b) it had to be paid out of pocket and the costs would not be reimbursed by health insurance or the government, (c) participants’ health and vaccination status permitted a booster vaccination, (d) participants could choose their preferred COVID-19 vaccine brand, and (e) the vaccination would be administered in a convenient location. Similar to Himmler et al. [34], we opted for a guided measurement procedure to help participants form their WTP for a non-marketed good. An initial filter question identified whether individuals were willing to pay a positive amount 1 3 1402 Does timing matter? The role of health information shocks in measuring willingness to pay for a booster vaccination. For those with a positive amount, we followed a three-step approach to elicit the WTP for a booster shot. First, we asked for the amount participants would certainly pay, using a scale ranging from €0 to €150 with visual anchoring points every €30. Alternatively, an open-ended question allowed participants to indicate a WTP greater than €150. Second, we asked for the amount participants would be unwilling to go beyond, using the same instruments as in the first step. Lastly, participants were asked to state their WTP using an open-ended question phrased in such a way that it reminded them of, and was conditional on, the previously set interval (see Online Resource). For participants who stated in the initial filter question that they had no WTP or a WTP of €0 in the third step, we asked for their motivation in order to differentiate between true zeros and protest answers that would indicate a violation of the hypothetical setting. Motivation options included (1) not needing the booster shot because the participant believed he or she would not become ill with COVID-19, (2) not being able to afford the booster shot, (3) the booster shot being of no value to the participant because of worries about potential side effects, (4) vaccines generally having no value to the participant, (5) not wanting to pay because of the belief that vaccines should be paid by the government, and (6) other reasons. Options 1 to 4 were considered to be reflective of a true WTP of €0. Options 5 and 6 were regarded as protest answers and thus excluded from the analysis. We excluded extreme values, defined as WTP values above the adjusted gross disposable income of households per capita for the year 2020 for each country [35]. We asked for valuations in participants’ local currency. For our analyses, we converted pound sterling and Danish krone to euros using the exchange rate of the European Central Bank from 23 December 2021. We adjusted WTP values for purchasing power parity based on the 2020 Eurostat purchasing power adjusted gross domestic product per capita [36]. We used conditional pathways, dynamic validation, and piped text to ensure the quality and consistency of answers. Lastly, we identified careless responders– e.g., those who completed the questionnaire in less than a third of the median survey duration per country (so-called speeders)– and dropped these observations [37]. Statistical analysis We calculated descriptive statistics for the sample, as well as WTP and HIS values. All analyses were based on the WTP response to the last elicitation step. We estimated the difference in WTP depending on the occurrence of HIS using a two-part model because of the continuous, nonnegative nature of the dependent variable WTP [38] and a large number of zeros. In the first part, we modeled the probability of the WTP being positive with logistic regression (WTP > 0; WTP = 0). In the second part, we estimated the WTP, conditional on it being positive, using a gamma distribution and log-link function. We considered four different model specifications for HIS: Model 1 included a binary variable indicating whether a participant had been subject to a HIS. Model 2 included a cardinal measurement of HIS intensity (i.e., the number of HIS experienced and its quadratic term). Model 3 referred to the most recent HIS experienced by a participant; it included the duration since the most recent HIS and the severity of the majority of COVID-19 cases. Lastly, Model 4 combined the specifications of Models 2 and 3 by adding the total number of HIS experienced in order to control for experience with COVID-19. We chose a linear rather than a squared specification for the number of HIS experienced by a participant because the results of Model 2 indicated that the tipping point was approximately six times the maximum reported number of HIS. Additionally, Model 4 controlled for the following predictors of WTP reported in the literature: socio-demographic characteristics, perceived threat from the disease, perceived benefits, and prior knowledge of the health intervention [39]. We operationalized socio-economic status using a binary variable for gender, six categories for age, and a three-level education variable based on each country’s educational qualifications (see Online Resource), as well as a four-category variable as a proxy for income indicating the extent to which participants were “able to make ends meet”, ranging from “easily” to “with great difficulty”. We operationalized the perceived threat from the disease as quality of life measured using EQ-5D-5 L and as subjective risk to own health from COVID-19 ranging from “no risk at all” to “very high risk”. In turn, we operationalized the perceived benefit of the health intervention by including (a) the vaccination status of the participant, measured using five levels ranging from not being vaccinated/not yet being vaccinated to having received up to three vaccination shots, and (b) the vaccination status of peers in four levels (“none”, “just a few”, “about half”, “most”). In addition, we included country fixed-effects and a risk-aversion measure based on Barsky et al. [40] with four levels ranging from “very low” to “high”. We chose this measure based on the assumption that it was independent of the COVID-19 pandemic and would be unaffected by the HIS examined in our study. We calculated average marginal effects (AME) to facilitate interpretation of the results of the two-part model in one estimate. We conducted the analysis using the “twopm” command [41] in Stata 17. 1 3 1403 C. Brinkmann et al. that among participants with a more recent HIS (5–6 weeks: €44.81, 7–8 weeks: €43.64, > 8 weeks: €47.57, Table 2). With regard to the degree of personal connection to COVID-19 cases, the differences in WTP among participants were small. Among those whose closest COVID-19 case was in the household, the WTP for a booster vaccination was €60.24, followed by a WTP of €55.99 if the closest case was in the family, €50.02 if it was among friends, and €47.64 if it was among colleagues. With regard to the severity of HIS, the mean WTP for those who reported that the severity of the majority of cases they heard of or experienced was milder than expected was €38.27, with those reporting higher severity also indicating a higher mean WTP (Table 2). Regression results The results of Model 1 show that HIS was associated with the WTP for a booster vaccination (Table 3), with the report of any HIS being associated with an increase in WTP of €14.54 compared to no HIS. In Model 2, our findings suggest that the total number of reported HIS and the square of this number were also associated with WTP, with the squared number of HIS indicating that an increasing number of HIS was associated with a decrease in WTP, but with a tipping point at approximately six times the maximally reportable number of HIS. In Model 3, which considered the duration since the most recent HIS, the association of a HIS with WTP was positive and stronger if the HIS occurred in the four weeks before the survey (AME for zero to two weeks: €20.05; AME for three to four weeks: €29.20) compared to no HIS (logistic: P <.003 for all temporal proximity levels, gamma: P =.0492 for reporting having experienced a HIS three to four weeks before the survey and all other levels P >.05). Moreover, WTP was positively associated with the severity of HIS: compared to the WTP among those who Results Descriptive results After we excluded careless responders, extreme values, protesters, and missing information, the final sample comprised 5809 observations (Fig. 1). Detailed sample characteristics can be found in Table 1. In total, 23.2% of participants expressed a WTP of €0, whereas 76.8% expressed a WTP greater than €0. Moreover, 61.9% of participants reported experiencing a HIS (i.e., at least one COVID-19 case), whereas 38.1% reported not experiencing any HIS. Of those who reported experiencing a HIS, 29.4% reported one, 22.0% reported two, 12.4% reported three, 10.6% reported four, and 14.2% reported five. The remaining 11.4% reported between six and the maximum number of 20 HIS. A large majority of participants who reported experiencing a HIS indicated that their most recent one had occurred zero to two weeks before the survey (69.5% of those who experienced a HIS), whereas 11.5% indicated that it had occurred three to four weeks and 9.9% that it had occurred more than eight weeks before the survey. On average, participants were willing to pay €48.58 for a booster vaccination. The WTP ranged from €0.00 to €7542.60. Participants who reported experiencing at least one HIS indicated having a higher WTP on average than those who reported not experiencing a HIS. The difference in the mean WTP between those with and those without a HIS varied across countries. With regard to the temporal proximity of a HIS, the mean WTP was €55.25 among individuals whose most recent HIS occurred in the two weeks before the survey and €60.62 among individuals whose most recent HIS occurred in the three to four weeks before the survey. Among individuals whose most recent HIS occurred more than four weeks before the survey, the WTP was lower than Fig. 1 Flowchart of the sample. HIS– Health information shock, WTP– Willingness to pay 1 3 1404 Does timing matter? The role of health information shocks in measuring willingness to pay expected. Other levels of severity were not associated with the WTP (Table 4). In Model 4, which included control variables, the significance and effect sizes mostly decreased for the duration since the most recent HIS and the severity of HIS. Compared reported the majority of cases as being as severe or as mild as expected, the WTP was lower among participants who reported the majority of COVID-19 cases as being milder (AME: €-19.81, logistic: P <.0001, gamma: P =.0248) than Table 1 Sample characteristics Sample characteristics per HIS status (N = 5809) Sample characteristics per WTP statement (N = 5809) No HIS (n = 2215) At least one HIS (n = 3594) WTP = 0 (n = 1349) WTP > 0 (n = 4460) n%n%n%n% Gender Female 1132 51.11 1899 52.84 804 59.60 2227 49.93 Male 1083 48.89 1695 47.16 545 40.40 2233 50.07 Age (years) 18–24 107 4.83 325 9.04 118 8.75 314 7.04 25–34 267 12.05 657 18.28 263 19.50 661 14.82 35–44 351 15.85 765 21.29 314 23.28 802 17.98 45–54 417 18.83 644 17.92 299 22.16 762 17.09 55–64 407 18.37 577 16.05 187 13.86 797 17.87 65 or older 666 30.07 626 17.42 168 12.45 1124 25.20 Education level High 867 39.14 1982 55.15 527 39.07 2322 52.06 Middle 944 42.62 1197 33.31 587 43.51 1554 34.84 Low 404 18.24 415 11.55 235 17.42 584 13.09 Ability to make ends meet Easily 350 15.80 570 15.86 115 8.52 805 18.05 Fairly easy 893 40.32 1543 42.93 418 30.99 2018 45.25 With some difficulty 780 35.21 1206 33.56 602 44.63 1384 31.03 With great difficulty 192 8.67 275 7.65 214 15.86 253 5.67 Country Denmark 223 10.07 542 15.08 161 11.93 604 13.54 France 409 18.47 421 11.71 250 18.53 580 13.00 Germany 388 17.52 308 8.57 173 12.82 523 11.73 Italy 218 9.84 470 13.08 147 10.90 541 12.13 Netherlands 254 11.47 414 11.52 213 15.79 455 10.20 Portugal 217 9.80 486 13.52 84 6.23 619 13.88 Spain 170 7.67 506 14.08 154 11.42 522 11.70 United Kingdom 336 15.17 447 12.44 167 12.38 616 13.81 Risk aversion Very low 684 20.66 1459 29.14 308 22.83 1312 29.42 Low 373 11.27 689 13.76 158 11.71 553 12.40 Moderate 261 7.88 555 11.08 130 9.64 459 10.29 High 1993 60.19 2304 46.02 753 55.82 2136 47.89 Vaccination status No 334 15.08 242 6.73 529 39.21 47 1.05 Not yet, but I intend to 35 1.58 61 1.70 55 4.08 41 0.92 Yes, the first shot 47 2.12 157 4.37 48 3.56 156 3.50 Yes, both shots 654 29.53 1432 39.84 459 34.03 1627 36.48 Yes, three shots (booster) 1145 51.69 1702 47.36 258 19.13 2589 58.05 Peers’ vaccination status Most 1731 78.15 2811 78.21 764 56.63 3778 84.71 About half 166 7.49 300 8.35 202 14.97 264 5.92 Just a few 211 9.53 417 11.60 271 20.09 357 8.00 None 107 4.83 66 1.84 112 8.30 61 1.37 1 3 1405 C. Brinkmann et al. 0.6625) compared to those who reported that the HIS was as severe or as mild as expected. Discussion Main findings When willingness-to-pay (WTP) exercises and other forms of contingent valuation are used to elicit population preferences, the timing of elicitation relative to a relevant event might be associated with the results. To investigate this issue, we analyzed whether recent health information shocks (HIS) were associated with WTP. The results of our regression analyses suggest that HIS are indeed positively associated with WTP for a booster vaccination for COVID-19, especially during the first four weeks after a HIS. This association may decrease over time but to reporting having not experienced any HIS, reporting that the most recent HIS had occurred in the two weeks or in the three to four weeks before the survey increased the WTP for a booster vaccination by €1.92 (logistic: P =.0025, gamma: P =.8060) and by €8.95 (logistic: P =.0095, gamma: P =.4427), respectively. These results suggest that having experienced a HIS was associated with being willingness to pay, but not with the height of the WTP (conditional on being positive). If the HIS was reported to have occurred between five and eight weeks before the survey, compared to reporting not having experienced a HIS, the AME decreased (€-4.60 and €1.80, respectively, all P >.05). Again, the severity of HIS was negatively associated with being willing to pay if the HIS was reported as having been milder than expected (€-6.14, logistic: 0.0093, gamma: 0.4451) and positively if it was reported as having been a bit more severe (€4.53, logistic: 0.0589, gamma: 0.6595) or more severe than expected (€5.23, logistic: 0.0842, gamma: Table 2 Mean WTP per country and HIS status HIS (n = 3594) No HIS (n = 2215) Overall (N = 5809) Mean (in €) SD (in €) Median (in €) Mean (in €) SD (in €) Median (in €) T-test HIS vs. No HIS Mean (in €) SD (in €) Median (in €) Overall 54.12 144.33 35.56 39.58 144.10 19.81 .0002 48.58 144.4 29.40 Country Denmark (n = 765) 66.30 76.55 51.62 45.91 60.81 33.89 .0001 60.36 72.87 43.01 France (n = 830) 41.13 42.50 29.72 37.46 296.83 11.89 .8042 39.32 210.43 19.81 Germany (n = 696) 70.51 78.41 52.29 51.77 92.39 29.05 .0039 60.06 86.92 40.67 Italy (n = 688) 49.89 60.75 35.56 33.68 58.29 17.78 .0010 44.75 60.41 31.11 Netherlands (n = 668) 74.92 376.42 37.71 28.50 35.45 18.86 .0131 57.26 297.86 31.43 Portugal (n = 703) 30.98 43.15 21.71 24.19 27.11 14.48 .0117 28.88 39.02 20.99 Spain (n = 676) 38.16 40.99 28.00 35.63 95.05 12.04 .7366 37.52 59.34 24.00 United Kingdom (n = 783) 68.72 117.19 47.04 48.06 94.21 29.40 .0064 59.85 108.35 35.28 Time since most recent HIS 0–2 weeks 55.25 161.88 36.13 3–4 weeks 60.62 115.27 43.01 5–6 weeks 44.81 60.64 32.53 7–8 weeks 43.64 42.48 31.43 More than 8 weeks 47.57 89.62 30.11 Degree of personal connection to COVID-19 cases Household 60.24 74.91 43.56 Family 55.99 210.26 36.16 Friends 50.02 67.63 34.54 Colleagues 47.64 65.86 34.67 Severity of majority of COVID-19 cases known by participant Milder than expected 38.27 62.62 24.00 A bit milder than expected 50.82 70.51 34.86 As severe/mild as expected 60.12 254.36 38.54 A bit more severe than expected 69.30 81.93 54.22 More severe than expected 66.74 61.50 57.90 Mixed, i.e., half of cases more severe, half of cases milder than expected 45.74 118.50 24.44 HIS– Health information shock, SD - Standard deviation 1 3 1406 Does timing matter? The role of health information shocks in measuring willingness to pay could nonetheless persist over a longer period. Our results also suggest that the association between HIS and WTP was especially pronounced through increasing the probability of having a positive WTP (as observed in the logistic regression), more so than with the height of the WTP conditional on being positive. When an individual receives news of a potential exposure in the recent past, it seems plausible that they might place a higher valuation on a preventive measure, such as a booster vaccination. However, we did not measure whether our participants indeed had contact with an infected person. Another explanation could be that participants might reevaluate the risk of developing COVID-19 themselves and the consequences of contracting it. The fact that the WTP associated with a HIS that occurred from two to four weeks before the survey was higher than the WTP associated with a HIS that occurred from zero to two weeks before the survey might be related to the time it takes for severe symptoms and their consequences to develop. Experiencing a health information shock, which can be perceived as a threat to one’s own health, might have a lasting consequence on individuals’ risk perceptions and thus their WTP preferences. This is supported by empirical findings from different fields. Evidence from the United States (US), for example, has shown that regional flood insurance purchases are highly correlated with the level of flood damage in the region during the prior year [42]. Similarly, Dave et al. (2020) examined risk perceptions of smoking e-cigarettes before and during an outbreak of e-cigarette or vaping-related lung injuries in the US in 2019 and 2020. They reported that risk perceptions decreased when participants received more information regarding the source of the outbreak, but not to pre-outbreak levels [24]. Other health shocks, like flu outbreaks, have been hypothesized to be associated with sustained changes in hygiene practices in developing countries [43]. Further evidence suggests similar longer-lasting effects [44]. Limitations Our study has a number of important limitations related to its setting, sample, measurement, and analysis that must be considered when interpreting its results. First, as we use cross-sectional data here, no causal claim can be made based on our results. Further, latent mechanisms driving both HIS and the booster intention might contribute to the associations observed in this study. For instance, individuals with a higher fear of COVID-19 might have tested themselves and their social environment more often for the virus, thus experiencing more HIS, than individuals with a lower fear of COVID-19. While the variable risk to own health attempts Table 3 Regression results model 1 and model 2 Model 1 Model 2 Logistic Gamma AME Logistic Gamma AME Est. SE p value 95%CI Est. SE p value 95%CI Est. Est. SE p value 95%CI Est. SE p value 95%CI Est. Intercept 0.74 (0.05) < 0.0001 0.65 , 0.83 4.07 (0.07) < 0.0001 3.94 ,4.20 0.80 (0.04) < 0.0001 0.72 , 0.88 4.01 (0.05) < 0.0001 3.90 ,4.12 HIS (Yes) 0.80 (0.06) < 0.0001 0.67 , 0.92 0.12 (0.08) 0.1493 -0.04 , 0.28 14.54 No. Of HIS 0.32 (0.03) < 0.0001 0.26 , 0.37 0.08 (0.03) 0.0077 0.02 , 0.14 7.26 No. Of HIS² -0.02 (0.00) < 0.0001 -0.02 , -0.01 0.00 (0.00) 0.1779 -0.01 , 0.00 -0.39 AME– Average marginal effect, Est.– Estimate, No.– Number, SE– Standard error 1 3 1407 C. Brinkmann et al. Model 3 Model 4 Logistic Gamma AME Logistic Gamma AME Est. SE p value 95%CI Est. SE p value 95%CI Est. Est. SE p value 95%CI Est. SE p value 95%CI Est. Intercept 0.74 (0.05) <.0001 0.65, 0.83 4.07 (0.06) <.0001 3.94, 4.19 1.57 (0.37) <.0001 0.84, 2.30 4.44 (0.31) <.0001 3.83, 5.05 No. Of HIS 0.12 (0.03) 0.0001 0.06, 0.17 0.03 (0.02) 0.0401 0.00, 0.07 2.34 Temporal proximity 0–2 weeks since HIS 1.04 (0.11) <.0001 0.83, 1.25 0.19 (0.11) 0.0953 -0.03, 0.41 20.05 0.50 (0.16) 0.0025 0.17, 0.82 -0.03 (0.10) 0.8060 -0.23, 0.18 1.92 3–4 weeks since HIS 1.03 (0.16) <.0001 0.71, 1.34 0.34 (0.17) 0.0492 0.00, 0.68 29.20 0.55 (0.21) 0.0095 0.14, 0.97 0.10 (0.13) 0.4427 -0.16, 0.36 8.95 5–6 weeks since HIS 0.86 (0.19) <.0001 0.48, 1.24 0.07 (0.22) 0.7437 -0.35, 0.50 12.20 0.39 (0.25) 0.1221 -0.10, 0.88 -0.16 (0.16) 0.3311 -0.47, 0.16 -4.60 7–8 weeks since HIS 0.82 (0.27) 0.0022 0.29, 1.34 0.02 (0.30) 0.9359 -0.56, 0.61 9.60 0.26 (0.33) 0.4365 -0.40, 0.92 0.00 (0.22) 0.9976 -0.42, 0.43 1.80 More than 8 weeks since HIS 0.80 (0.16) <.0001 0.48, 1.11 0.12 (0.18) 0.5106 -0.24, 0.48 14.09 0.22 (0.20) 0.2882 -0.18, 0.61 0.02 (0.13) 0.8997 -0.25, 0.28 2.30 No HIS . . . . . Severity of HIS Milder than expected -0.70 (0.12) <.0001 -0.93, -0.46 -0.31 (0.14) 0.0248 -0.57, -0.04 -19.81 -0.41 (0.16) 0.0093 -0.71, -0.10 -0.08 (0.10) 0.4451 -0.27, 0.12 -6.14 A bit milder than expected -0.05 (0.14) 0.7283 -0.32, 0.22 -0.16 (0.14) 0.2297 -0.43, 0.10 -7.99 -0.03 (0.17) 0.8767 -0.35, 0.30 0.01 (0.10) 0.8871 -0.18, 0.21 0.53 As severe/mild as expected . . . . . A bit more severe than expected 0.21 (0.16) 0.1852 -0.10, 0.53 0.12 (0.15) 0.4333 -0.17, 0.40 8.18 0.36 (0.19) 0.0589 -0.01, 0.73 0.05 (0.11) 0.6595 -0.16, 0.26 4.53 More severe than expected 0.09 (0.18) 0.6228 -0.27, 0.45 0.10 (0.17) 0.5744 -0.24, 0.44 5.98 0.39 (0.23) 0.0842 -0.05, 0.83 0.06 (0.13) 0.6625 -0.20, 0.32 5.23 Mixed, i.e. half more severe, half milder than expected -0.42 (0.18) 0.0193 -0.78, -0.07 -0.23 (0.20) 0.2649 -0.62, 0.17 -13.84 -0.40 (0.23) 0.0798 -0.84, 0.05 -0.06 (0.15) 0.6679 -0.35, 0.23 -5.48 No HIS . . . . . Female 0.20 (0.08) 0.0180 0.03, 0.36 -0.02 (0.06) 0.7070 -0.13, 0.09 0.19 Age 18–24 years old -0.43 (0.19) 0.0245 -0.81, -0.06 0.22 (0.14) 0.1067 -0.05, 0.48 9.36 25–34 years old -0.65 (0.16) <.0001 -0.96, -0.35 0.02 (0.11) 0.8248 -0.18, 0.23 -2.63 35–44 years old -0.64 (0.15) <.0001 -0.93, -0.35 0.02 (0.10) 0.8023 -0.16, 0.21 -2.48 45–54 years old -0.78 (0.14) <.0001 -1.05, -0.50 -0.01 (0.09) 0.9420 -0.18, 0.17 -4.87 55–64 years old -0.24 (0.15) 0.1185 -0.53, 0.06 -0.04 (0.09) 0.6790 -0.21, 0.13 -2.95 65 years old and older . . Education level High 0.12 (0.13) 0.3525 -0.13, 0.37 0.15 (0.09) 0.0940 -0.03, 0.33 7.67 Table 4 Regression results model 3 and model 4 1 3 1408