Individual characteristics associated with perceptions of control over mortality risk and determinants of health effort

People who believe they have greater control over health and longevity are typically more likely to invest in their long‐term health. Investigating individual differences in perceived control over risk and exploring different determinants of health effort may help to tailor health promotion programs to more effectively encourage healthy behaviors. From a sample of 1500 adults, we measured perceived control over 20 causes of death, overall perceived uncontrollable mortality risk (PUMR), state‐level optimism, self‐reported health effort, and the accuracy of estimations of avoidable deaths. We found individual differences in perceptions of control over specific causes of death based on age, gender, and income. PUMR was predicted by socioeconomic variables expected to influence exposure to risk and resource availability. Higher levels of PUMR, not perceptions of control over specific causes of death, predicted self‐reported health effort. The strength of relationship between PUMR and lower health effort was not moderated by state‐level optimism. Age and education both positively predicted greater accuracy in assessing the prevalence of avoidable deaths. We suggest that PUMR may capture people's “general sense” of mortality risk, influenced by both exposure to hazards and the availability of resources to avoid threats. Conversely, perceived control over specific risks may involve more deliberate, considered appraisals of risk. This general sense of risk is thought to play a more notable role in determining health behaviors than specific assessments of control over risk. Further study is needed to investigate the degree to which PUMR accurately reflects objective measures of individual risk.


INTRODUCTION
Understanding individual differences in perceptions of control over health and longevity is important because people who believe they have greater control are typically more likely to invest in their long-term health (Grotz et al., 2011;Leganger & Kraft, 2003;Norman et al., 1998).Investigating the individual characteristics associated with perceptions of control over health and longevity may help to promote positive health behaviors among those with a low sense of control.Previous research has provided a variety of theories to account for differences in perceived control and the subsequent influence on health behaviors.For example, much of the literature has investigated differences in perceived control over health by measuring health locus of control (Cheng et al., 2016;Fischer et al., 2006;Wallston, 2005;Wallston et al., 1978).Health locus of control captures the degree to which people attribute the state of their own health to factors either within or beyond their personal control (Cheng et al., 2016;Wallston, 2005;Wallston et al., 1978).For example, previous studies have consistently reported age-, gender-, and socioeconomic-based differences in the extent to which Risk Analysis.2024;44:1339-1356.
wileyonlinelibrary.com/journal/risa 1339 people attribute the state of their current and future health to different dimensions of control (Deroche et al., 2012;Grotz et al., 2011;Poortinga et al., 2008;Specht et al., 2013).Men typically report greater levels of perceived internal control over their health, and women typically report greater levels of perceived external control (Poortinga et al., 2008;Specht et al., 2013).Older age and lower socioeconomic status have also been associated with a reduced sense of internal control over health outcomes (Grotz et al., 2011;Poortinga et al., 2008).Health locus of control has largely been defined as an individual difference variable, more "trait-like" than "statelike" (Galvin et al., 2018;Rotter, 1966;Wallston, 2005).Furthermore, it has been repeatedly suggested that this trait mediates the relationship between lower socioeconomic status and unhealthy behaviors (Jang & Baek, 2018;Park et al., 2018;Pedron et al., 2021;Stephenson-Hunter & Dardeck, 2019).However, the literature on health locus of control does not provide a plausible account for why there should be persistent individual differences in the extent to which health is considered controllable.In response to this, we investigate the individual characteristics associated with perceptions of control over mortality risk in light of a theoretical account that does offer an ultimate explanation for social gradients in health, the uncontrollable mortality risk hypothesis (UMRH).
The UMRH provides an ultimate explanation for socioeconomic gradients in health by modeling the expected health response to experiencing higher levels of uncontrollable risk.The UMRH suggests that people who are more likely to die as a result of factors that are beyond their control should be less motivated to engage in positive health behaviors (Brown & Pepper, 2022, 2023;Pepper & Nettle, 2014a).This is based on a behavioral ecological model for explaining social gradients in preventative health behaviors, which states that the optimal individual investment in health behavior should be less for people of lower socioeconomic status (Nettle, 2010).It argues that human psychological mechanisms respond to the presence of environmental cues of risk to determine the optimal level of investment in preventative health behaviors.People in lower socioeconomic positions typically experience greater exposure to environmental risks that are harmful to their health (Bolte et al., 2010;Cifuentes et al., 2021;Evans & Kim, 2010).Pepper and Nettle (2014b) found that the effect of lower socioeconomic status on reported health effort was mediated by perceived uncontrollable mortality risk (PUMR), suggesting that socioeconomic variation in health behavior may therefore be partly attributable to exposure to risks beyond individual control.Additionally, recent research during the COVID-19 pandemic found that PUMR was associated with lower reported adherence to advice on diet and physical activity, as well as higher levels of smoking (Brown et al., 2021a(Brown et al., , 2021b)).Despite this evidence, the socio-demographic determinants of PUMR and their impact on health behaviors have been understudied.It has been argued that sociodemographic variables, such as gender, age, income, and education, are rarely the central focus of risk perception research and are typically only included as control variables (Siegrist & Árvai, 2020).Given the link between increased levels of PUMR and lower reported health effort, investigating the individual characteristics and sociodemographic determinants of PUMR may help to identify people whose health behavior is most likely to suffer as a result of their perceptions of risk.
Further study is also needed to better understand the reported negative relationship between PUMR and health effort.In addition to the sociodemographic determinants of perceived control over health, it is possible that certain psychological factors influence the strength of association between PUMR and preventative health behaviors.For example, optimism has been found to predict a range of health perceptions, behaviors, and outcomes.In a recent review of health literature, optimism was consistently associated with a lower risk of developing cardiovascular disease and related mortality risk (Amonoo et al., 2021).The same review suggested that this relationship persists even when controlling for various sociodemographic and psychological factors (such as income and depression) consistently associated with health outcomes.The relationship between optimism and health outcomes may be explained by the positive association between optimism and preventative health behaviors, such as physical activity, smoking cessation, diet, and medication adherence (Amonoo et al., 2021;Boehm et al., 2018;Hingle et al., 2014).However, the causal direction of the association between optimism and health is hard to disentangle (Chopik et al., 2015).People in a good state of health may be more likely to feel optimistic than those who are unwell.Conversely, feeling more optimistic may make someone more likely to engage with their health.Optimism has been consistently positively associated with perceptions of control (Klein & Helweg-Larsen, 2002).However, like optimism, much of the literature on perceived control has conceptualized the construct as an individual difference variable, more "trait-like" than "state-like" (Galvin et al., 2018;Rotter, 1966;Wallston, 2005).Shanahan et al. (2020) found that levels of individual optimism predicted participants' expectations regarding the outcome of hypothetical situations when presented with an uncontrollable situation.Furthermore, optimism was found to moderate the relationship between external locus of control and health and wellbeing among students, such that the negative relationship between external locus of control and college-related well-being is weakened among those who are more optimistic (Cuze & Aleksic, 2021).Given the suggested association between optimism and perceptions of control, as well as the established influence of both constructs on health behaviors, it is possible that levels of optimism may moderate the negative relationship between PUMR and health effort.For example, the negative association between PUMR and health effort may be weakened for those who are typically more optimistic than others.Similarly, this relationship could be strengthened among those who are lower in optimism.Investigating the influence of psychological constructs such as optimism on the relationship between PUMR and health effort may help to better understand some of the individual differences likely to determine the impact of perceived control on health effort.
Finally, people differ in their ability to accurately assess the levels of risk around them.Understanding individual characteristics likely to impact the accuracy of risk estimations is important because of the discussed relationship between perceptions of risk and subsequent health behaviors.For example, those who underestimate the prevalence of certain risks may be less motivated to engage in relevant protective health behaviors (Burns, 2005).This has implications for public health policy because interventions that seek to reduce levels of exposure to certain risks in order to enhance public perceptions may be ineffective in encouraging preventative health behaviors if people do not accurately perceive improvements to the safety of their environment.Investigating the individual characteristics associated with the accuracy (or inaccuracy) of perceived risk may help to understand who is most likely to inaccurately assess their own level of risk.This may lead to health interventions or targeted informational campaigns aimed at aligning perceptions with objective levels of exposure in order to encourage individuals to appropriately respond to risk.
By investigating individual differences in perceived control over risk, and by evaluating different determinants of health effort, we hope to inform health promotion efforts by elucidating the characteristics most likely to influence health perceptions and behaviors.The specific aims of this study are as follows: 1. Investigate the individual characteristics that predict levels of perceived control over mortality risk.We will investigate socioeconomic and demographic drivers of perceived control over specific causes of death, as well as drivers of overall levels of PUMR.We will then consider whether perceived control over specific risks and overall PUMR are predicted by the same, or different, individual characteristics.2. Investigate the relationship(s) between perceived control over mortality risk and health effort.We will determine whether or not perceptions of control over multiple individual causes of death predict reported health effort and how this relationship compares to that between PUMR and health effort.3. Explore the potential role that optimism plays in moderating the negative relationship between PUMR and health effort in order to better understand the effects of PUMR. 4. Explore the influence of individual characteristics on the accuracy of risk estimations of the prevalence of avoidable deaths in the United Kingdom.  2 and 3 for full sample characteristics).

Personal information
Measuring socioeconomic status in health research is fraught with difficulties due to the complex nature of the construct and the lack of precision and reliability of measures seeking to capture it (Shavers, 2007).Recent guidance for measuring socioeconomic status has highlighted the importance of selecting those structural features of socioeconomic status most central to the outcomes under investigation (Antonoplis, 2022).This led us to consider which socioeconomic features might reasonably be expected to influence perceptions of control over risk.For example, socioeconomic factors relevant to the degree of exposure to risk experienced in the home and working environment may impact perceptions of control over risk.Measuring beliefs about neighborhood safety and levels of exposure to occupational hazards may capture some of these socioeconomic features.Additionally, income might represent a socioeconomic component likely to influence the extent to which people feel they can mitigate or avoid certain risks.Subjective measures of the degree to which people feel their income is sufficient for meeting the demands of daily life may be of particular significance for studying beliefs about the control of long-term health and mortality.In consideration of the above socioeconomic features expected to influence perceptions of control over risk, we include measures of occupational exposure to risk, perceived neighborhood safety, subjective and objective measures of discretionary income, and education to capture a broad range of variables relevant to socioeconomic status.
First, participants provided their age, gender, and number of years in post-16 education.Participants answered a single-item measure for household discretionary income to capture the money they have left each month after paying for all household necessities.Participants were given instructions for how to calculate their household discretionary income and provided their score as a free text response in pounds sterling, which was then divided by the number of people Perceived neighborhood safety Perceived neighborhood safety measured on a seven-point scale (An et al., 2017;Prins et al., 2013) Occupational exposure to risk Physical working environment subscale from the European Working Conditions Survey 2015 (Eurofound, 2017).Reported exposure to 13 physical conditions of employment on a seven-point Likert scale from "never" to "all of the time" Numeracy 10-Item numeracy measure (Lipkus et al., 2001), responses to which are summed to give a score from 0 to 10 Perceived uncontrollable mortality risk (PUMR) PUMR measure (Pepper & Nettle, 2014b) measured on a continuous scale from 0 to 100 Perceived control of multiple causes of death Individual scores for perceived control over a list of causes of death (see Table S1), measured on a continuous scale from 0 to 100 Accuracy of perceived risk relative to objective measures of mortality risk Objective measure of risk a minus corresponding perceived prevalence of avoidable deaths Overall state optimism score Validated seven-item measure of state optimism (Millstein et al., 2019), measured on a five-point Likert scale from "strongly disagree" to "strongly agree" Current health effort Self-reported current investment in ensuring one's health and safety (Pepper & Nettle, 2014b) measured on a continuous scale from 0 to 100 a The objective measure of risk is the percentage of avoidable deaths in the United Kingdom accounted for by each category of avoidable death (Office for National Statistics, 2021a).& Wells, 1989;Rader et al., 2011).This scale captured the extent to which participants agreed with three statements about whether or not they felt their household finances could satisfy their needs.Participants indicated the extent to which they agreed with the statements, "My household income is high enough to satisfy nearly all my important desires" and "My household has more to spend on extras than most of my neighbours do," on five-point Likert scales ranging from strongly disagree to strongly agree.The third survey item was reverse-scored, "No matter how fast my household income increases, I never seem to get ahead."The value for Cronbach's α for the three-item SDI scale was α = 0.64, indicating an "adequate" degree of internal consistency (Taber, 2018).

TA B L E 2
The three survey items were summed to produce a single score for SDI.

Perceived neighborhood safety
Participants answered an existing measure of perceived neighborhood safety to capture the extent to which they feel safe in their home surroundings (An et al., 2017;Prins et al., 2013).Participants indicated the extent to which they agreed with the statement, "I feel safe in my neighbourhood" on a seven-point Likert scale ranging from strongly disagree to strongly agree.

Occupational exposure to risk
Participants currently in employment (n = 999) provided a measure of occupational exposure to risk taken from the European Working Conditions Survey (Eurofound, 2017).
The physical environment subscale was used to capture the extent to which participants feel they are exposed to a range of potentially harmful factors in the workplace.Participants answered 13 survey items for workplace hazards, indicating their frequency of exposure to each risk on a seven-point Likert scale ranging from never to all of the time.The survey items included occupational risks, such as maintaining tiring or painful positions, carrying or moving heavy loads, and being exposed to hazardous materials, high temperatures, or noisy environments.The value for Cronbach's α for the 13item physical environment subscale was α = 0.84, indicating a "good" degree of internal consistency (Taber, 2018).The 13 survey items were summed to produce a single score for occupational exposure to risk.

Numeracy
An existing measure of numeracy was used to capture basic numerical understanding (Lipkus et al., 2001).Participants were asked 10 numerical questions, such as "Imagine that we rolled a fair, six-sided dice 1000 times.Out of 1000 rolls, how many times do you think the dice would come up even (two, four, or six)?" (Answer: 500 out of 1000).Participants were scored as either correct or incorrect for each of the 10 numerical questions.The value for Cronbach's α for this numeracy measure was α = 0.65, indicating an "adequate" degree of internal consistency (Taber, 2018).The 10 question responses were summed to produce a single score for numeracy (0 indicating that all responses were incorrect; 10 indicating that all responses were correct).This measure of numeracy was included because asking participants to numerically express their perceived degree of control or likelihood of risk has repeatedly been found to create challenges for accurately capturing beliefs (Peters, 2008;Rothman & Kiviniemi, 1999).Raude et al. (2023) recently reported that the magnitude of the consistently reported primary bias in risk perception (in which people typically overestimate rare risks to their health and underestimate common risks; Hakes & Viscusi, 2004;Lichtenstein et al., 1978;Slovic, 1978) varies as a function of the respondents' level of numeracy.We will use this measure to account for the role that individual numeracy plays in potential perceptual biases of risk when estimating the prevalence of different causes of death.

Perceptions of risk
Perceived control over specific causes of death was measured by asking, "how much do you feel you could control your own risk of dying from each of the following causes if you did everything you could to take care of your health and ensure your safety?" for 20 different causes of death.These causes consisted of the Office for National Statistics' 7 most common categories of avoidable deaths in the United Kingdom (including heart disease, cancers, and drug-and alcoholrelated death), and 13 additional risks considered most serious and likely from the UK National Risk Register 2020 and highlighted in recent qualitative interviews about perceptions of control over risk (including risks from pollution, contaminated food, and occupational hazards); see Table S1 for a full list of causes of death (Brown et al., 2022;Cabinet Office, 2020; Office for National Statistics, 2021a).Scores for perceived control over the cause of death were provided on a sliding scale from 0 "very unlikely" to 100 "very likely."Participants provided scores for overall levels of PUMR.Using the PUMR measure devised by Pepper and Nettle (2014b), participants were asked, "If you were to do absolutely everything you could to take care of your health and ensure your safety, what do you think the chances would be that you would live to be [UK life expectancy for their gender] or more?"Current life expectancy at birth in the United Kingdom is 79 for males and 83 for females; participants who reported a different gender identity were shown the population average of 81 years.Participants responded on a sliding scale from 0 no chance to 100 certain.Responses were subtracted from 100 to provide PUMR scores, representing that portion of mortality risk, which the participant believes is beyond their control.
Perceived prevalence of risk was measured by asking participants to estimate the percentage of avoidable deaths in the United Kingdom caused by each of the ONS' seven categories of avoidable death (infectious diseases, cancers, diseases of the circulatory system, diseases of the respiratory system, injuries, alcohol-and drug-related deaths, and other causes of avoidable death) (Office for National Statistics, 2021a).Cumulative scores for all causes had to equal 100% of UK avoidable deaths.
Accuracy of perceived risk prevalence was captured by measuring the absolute difference between scores for how much participants felt each cause of death accounted for avoidable deaths in the United Kingdom and an objective measure of mortality prevalence.This objective measure was the percentage of deaths accounted for by each category of avoidable death in 2020 (Office for National Statistics, 2022).

State optimism
Participants provided scores for a validated seven-item measure of state-level optimism (Millstein et al., 2019).
Participants indicated their level of agreement with seven statements about their current state of optimism.For example, participants indicated the extent to which they agreed with the statement, "At the moment, I expect more to go right than wrong when it comes to my future" on a Likert scale from one strongly disagree to five strongly agree.The value for Cronbach's α for the seven-item measure of state-level optimism was α = 0.94, indicating an "excellent" degree of internal consistency (Taber, 2018).The seven survey items were summed to produce a single score for state-level optimism.A measure of state optimism was chosen in preference to trait optimism in order to measure levels of participant optimism at the time of providing scores for perceptions of risk rather than measuring a more general propensity toward optimism over time.However, the previously published validation of this measure reported that state and trait optimism are highly correlated (Millstein et al., 2019).

Health effort
Current health effort was recorded using a single-item measure taken from previous research into the relationship between PUMR and reported investment in health (Pepper & Nettle, 2014b).Participants were asked, "How much effort do you make to look after your health and ensure your safety these days?" on a scale from 0 no effort at all to 100 the maximum effort you could make.
BestNormalize performs a suite of normalizing transformations and then selects the transformation that has the lowest Pearson P test statistic for normality.Data transformation was used in preference to conducting nonparametric tests because previous research has suggested that this can yield more accurate parameter estimates in regression analysis and improved statistical power (Osborne & Waters, 2002;Tabachnick et al., 2013).Our full R script is available alongside our preregistration [osf.io/dgwna].
A multivariate multiple regression analysis was conducted to investigate the impact of socioeconomic, demographic, and occupational exposure variables on perceptions of individual control over different causes of death.This analysis tested the influence of socioeconomic status and additional covariates on perceived control over 10 high-level categories of causes of death (see Table S1).
For analyses containing the variable PUMR, participants were excluded if they had already reached the age of average life expectancy for their gender (n = 2).This is because the PUMR measure asks participants to indicate how likely they believe they are to reach the average life expectancy.Overall accuracy scores for estimations of the prevalence of avoidable deaths across all categories were calculated for each individual participant by summing the absolute distance between perceived prevalence scores and the objective prevalence of each category of avoidable death.
The data associated with this study were also used to produce a second article, "Perceptions of control over different causes of death and the accuracy of risk estimations" (Brown et al., 2023), which describes differences between causes of death with respect to perceived control, perceived personal likelihood of death, certainty of risk estimation, perceived knowledge, and the accuracy of perceived prevalence of risk.

Individual predictors of perceived control over different causes of death
Multivariate multiple regression analysis was used to investigate the impact of socioeconomic, demographic, and occupational exposure variables on perceptions of control over 10 causes of death (see Table 4).Age, gender, and SDI predicted perceptions of control over multiple causes of death.Multivariate analysis of variance was conducted to evaluate the overall significance of predictor variables across the multiple outcome variables.Age (Pillai's trace = 0.12, F 10,981 = 12.83, p < 0.001), gender (being male; Pillai's trace = 0.07, F 10,981 = 7.13, p < 0.001), and SDI (Pillai's trace = 0.04, F 10,981 = 3.78, p < 0.001) each showed a significant positive overall effect on levels of perceived control over the 10 causes of death (see Table 5).
Examining each dependent variable from the multivariate multiple regression separately, age was a significant predictor of increased perceived control over 9 out 10 of the causes of death (the exception being risk of death due to antimicrobial resistance).To visualize the effect of age on perceived control of specific risks, we split our sample at the median age to observe differences in perceptions of control between over 45 s and those aged 45 or below.Older participants reported greater control over their risk of dying due to 9 out of 10 causes when compared to younger participants (see Figure 1 and Table 4).Furthermore, a visual inspection of the data indicates a linear relationship between age and overall perceived control over specific causes of death (see Figure S1).
Follow-up inspection of the effects of socioeconomic and demographic predictors on individual causes of death showed that the global effect of gender on perceptions of control was driven by 2 out of 10 risks.Gender (being a man) was a significant predictor of perceived control over death due to injuries and catastrophic risks only.For these two risks, men reported greater perceived control than women (see Figure 2 and Table 4).
The global effect of SDI on perceived control over risks was due to SDI being a significant predictor of 4 out of 10 causes of death (cancer, respiratory disease, environmental risks, and catastrophic risks).To visualize this finding, we split our sample by the median point for SDI and categorized these groups into high and low SDI.Those whose scores for SDI were higher than the median response reported greater perceived control over risk of death due to cancer, respiratory disease, environmental risks, and catastrophic risks (see Figure 3 and Table 4).

TA B L E 4
Multivariate multiple regression results showing standardized regression weights for predictors of perceived control over specific causes of death.

Determinants of health effort
The mean score for reported health effort was 67.43 out of a possible 100 (SD = 19.62)indicating that people, on the whole, believe they are making approximately two thirds of the maximum effort they could make to take care of their health.Though the association was conventionally a small one, PUMR was found to be a better predictor of self-reported health effort than perceptions of control over individual causes of death (see Table 6).From a linear regression model containing perceptions of control over each cause of death, and controlling for socioeconomic and demographic variables, only drugs and alcohol (β = 0.09, 95% CI [0.04, 0.15], p < 0.01) and antimicrobial resistance (β = 0.09, 95% CI [0.03, 0.14], p < 0.01) predicted self-reported health effort (trivial effect sizes [Lovakov & Agadullina, 2021] (see Table 6 and Tables S3 and S4 for correlations between proposed predictors of health effort and analysis, which indicate no collinearity between predictors [Daoud, 2017]).When PUMR was added to the model, model fit increased and higher scores for PUMR predicted lower health effort (small effect size [Lovakov & Agadullina, 2021]; β = −0.20,95% CI [−0.14,−0.25],p < 0.01; see Table 6).
The mean score for state optimism was 23.03 of a possible 35 (SD = 6.56).Pearson correlation tests showed that state optimism was significantly associated with scores for both PUMR (r = −0.30,p < 0.01) and reported health effort (r = 0.32, p < 0.01; for further correlations among state optimism, PUMR, and socioeconomic, demographic, and occupational exposure variables, see Table S5).In a linear regression model, it was found that state optimism did not moderate the negative association between PUMR and reported health effort, as there was no significant interaction effect between PUMR and optimism as predictors of reported health effort (β = −0.01,95% CI [−0.05, 0.04], p = 0.80; see Table S6).

DISCUSSION
This study found age-, gender-, and income-based differences in perceptions of control over individual causes of death.Levels of PUMR were driven by income and perceived neighborhood safety.For the most part, perceptions of control over specific risks did not predict self-reported health effort (with the exception of trivial effects for drugs and alcohol and antimicrobial resistance).However, higher levels of PUMR did predict lower self-reported health effort.The strength of relationship between PUMR and health effort was not moderated by levels of state optimism, despite higher optimism also predicting greater self-reported health effort.Finally, the age and number of years in post-16 education both positively predicted greater accuracy when estimating the prevalence of UK avoidable deaths.

Individual differences in perceptions of control over risk
Our findings show differences in perceived control over different causes of death based on age, gender, and SDI.Men reported feeling that they have greater control over some risks compared to women, but only with respect to the risk of death due to injuries and catastrophes.Similarly, there was a small effect of higher SDI on greater perceived control for just under half of all causes of death (cancer, respiratory disease, environmental risks, and catastrophic risks).More notably, older participants felt they had more control over their risk of dying due to 9 out of 10 causes when compared to younger participants.Previous research has found that perceptions of control over health can differ over the lifespan with perceived control increasing in early adulthood and dropping in later life (Bailis et al., 2010;Mirowsky, 1995;Sargent-Cox & Anstey, 2015).However, our data indicate a positive linear relationship between age and perceived control over different causes of death, whereby participants in the highest age-brackets did not report lower levels of control.It is possible that this is because our participants were recruited online (via Prolific) and therefore must have had Internet access and a sufficient degree of computer literacy to participate in the study.Previous research has highlighted the potential impact of online recruitment methods on the generalizability of findings from online studies of older adults (Remillard et al., 2014).The individual characteristics associated with overall PUMR scores differed from those driving perceived control over different causes of death.For example, there were no differences in PUMR with respect to age or gender, but there were clear differences for perceptions of control over specific causes of death.Instead, PUMR was associated with socioeconomic variables that we may theoretically expect to influence an individual's perceived exposure to risk; PUMR was negatively predicted by perceived neighborhood safety and income.We might reasonably expect that feeling that you live in an unsafe neighborhood, especially if you are unable to afford to live elsewhere, may generate greater PUMR.Higher levels of income are also likely to steer someone to think they have greater ability to mitigate the risks they experience.A wide range of factors can influence the ways in which income inequality affects health behaviors and subsequent outcomes (Bor et al., 2017).Those with a higher income are more likely to be able to make adjustments to the safety of their home and select the environment in which they live.Greater income is also likely to afford improved access to healthier foods, wider options for medicines and treatments, as well as additional tools to navigate essential services.Perceived neighborhood safety and income were captured to investigate the relationship between socioeconomic status and perceptions of control over risk.Our findings show that lower socioeconomic status is associated with greater PUMR across these factors.This further highlights the importance of selecting structural features of socioeconomic status that are theoretically relevant to key variables of interest (Antonoplis, 2022).These findings suggest that PUMR may represent some psychological appraisal of exposure to risk in one's local environment, as well as one's perceived means to mitigate risk.This contrasts with perceived control over different causes of death, for which age was found to be the key driver of perceptions.Instead, we propose that these socioeconomic diversities of PUMR suggest that the construct reflects a general sense of one's environment that may be influenced by both perceived exposure to risk and the resources at someone's disposal to avoid threats to health and longevity.Future research may look to investigate whether this general sense of risk, captured by PUMR, relates to the concept of ambiguous risk described in the Safety Perceptions Index (SPI) from the Lloyd's Register Foundation (2023).Ambiguous risk refers to people's general sense that risk exists in the world around them but cannot be precisely defined.The SPI 2023 reported a rise in ambiguous risk between 2019 and 2021.The SPI 2023 suggests that the rise in ambiguous risk may have resulted from an increase in generalized and indeterminate fear in response to the COVID-19 pandemic, as well as a changing media landscape that has an increased tendency to amplify misleading or inconclusive information.

Determinants of health effort
In addition to differences between PUMR and perceived control over different causes of death in terms of the individual characteristics driving perceptions, there were also differences in the degree to which they predict health effort.With the exception of one small and one trivial effect (for the risk of death due to drugs and alcohol and antimicrobial resistance, respectively; (Lovakov & Agadullina, 2021)), perceptions of control over individual causes of death did not predict self-reported health effort.However, in the support of previous research (Brown et al., 2021a;Pepper & Nettle, 2014a, 2014b), higher levels of PUMR did predict lower self-reported health effort.This indicates a difference in the effect of perceptions between individual assessments of control over specified risks and the discussed general sense of one's experience of risk captured by PUMR.It is suggested that this difference may be akin to an existing division in the risk literature drawn between risk as analysis and risk as feelings (Slovic, 2020;Slovic et al., 2004).Risk as analysis describes the deliberate, logical, and considered cognitive assessment of risk.Risk as feelings captures intuitive, affective, and automatic perceptual responses to risk.We argue that people may respond to questions about control over specific causes of death by adopting a risk as analysis approach, thus calculating their perceived level of control through deliberate assessment.However, our measure of PUMR may be more likely to capture people's general sense of control over mortality risk by triggering a more instinctive response.
It has been proposed that behavioral health responses to cues of risk are more likely to rely on automatic perceptual responses rather than slower cognitive processes (van't Riet & Ruiter, 2013).This is because, for much of human history, we have been exposed to unstable and threatening environments.Therefore, it is more likely that evolution would have favored psychological mechanisms providing faster, instinctive responses to cues of risk, in preference to comparatively slower evaluative appraisals of threat (van 't Riet & Ruiter, 2013).That said, it is unlikely that the degree to which someone invests in preventative health behaviors should be entirely reliant on automatic perceptual processes.A central debate within contemporary research into the drivers of health habits is the extent to which health behaviors are controlled by automatic psychological mechanisms (Hagger, 2020).Researchers have argued that complex behaviors (such as those required for engaging with preventative health) depend on dual-process models of action and should not simply be reduced to acting either deliberately or automatically (Gardner et al., 2016;Phillips, 2020).A habit associated with engaging in preventative health behavior may be defined as "a behavioral tendency that is enacted with little conscious awareness or reflection, in response to a specific set of associated conditions or contextual cues" (Hagger, 2020).Though the pathways between environmental cues and preventative health behaviors are varied and complex, our results indicate that someone's general sense of the uncontrollability of risk in their environment may play a notable role in determining individual investment in health behaviors.

Optimism and PUMR have separate effects on health effort
PUMR and state-level optimism both predicted self-reported health effort.PUMR and optimism were negatively associated with one another and were both predicted by SDI and neighborhood safety.However, there was no statistical interaction effect between PUMR and optimism on reported health effort, indicating independent effects.First, our findings support the previously reported positive association between optimism and a range of preventative health behaviors, such as physical exercise, quitting smoking, following a healthy diet, and treatment adherence (Amonoo et al., 2021;Boehm et al., 2018;Hingle et al., 2014).The negative relationship between state-level optimism and PUMR was such that participants who were more optimistic at the time of the survey reported lower levels of PUMR.However, optimism was found not to moderate the negative relationship between PUMR and health effort.We therefore propose that higher levels of optimism are unlikely to have a protective effect for those who perceive their exposure to uncontrollable risk as high, in that being optimistic would not diminish the strength of the effect of PUMR on negative health behaviors.Similarly, it is unlikely that being low in optimism would worsen the effects of PUMR on health behaviors.Previous research has reported that optimism has unique predictive effects on health over time that are separate from various psychological traits, mental health conditions, and subjective appraisals of well-being (Carver, 2014;Chopik et al., 2015).It is possible that the strength of the influence of PUMR on health efforts may be influenced by psychological factors separate from optimism.These could include personality traits, implicit attitudes, or beliefs concerning the severity or immediacy of the threats we face (Hagger, 2020).However, given that PUMR is predicted by those socioeconomic variables capturing exposure to risk and resource availability, further research should first assess the environmental triggers of perceived control rather than exploring additional psychological characteristics.

Accuracy of risk estimations
Age and the number of years in post-16 education both positively predicted greater accuracy when assessing the prevalence of UK avoidable deaths.This indicates that being older and more educated may improve estimations of risk.This offers support to previous research, which has argued that estimations of the prevalence of different causes of death may be more accurate for those who are older and more educated (Andersson, 2011;Hakes & Viscusi, 2004).Accurately perceiving risks can help us to respond appropriately to the threats we face and to direct preventative action (Knuth et al., 2015).We propose that public health interventions should look to use informational campaigns targeting younger, less educated members of society with an aim to aligning perceptions with objective levels of risk.This may help to motivate individuals to respond appropriately to key risks to their health and longevity.Given the much discussed relationship between increased levels of PUMR and lower health effort, further research is needed to explore the extent to which individual scores for PUMR are associated with objective measures of individual risk.Isch et al. (2023) provided initial support for the notion that PUMR maps somewhat onto environmental levels of objective risk.However, individuals are exposed to threats to health and longevity through a myriad of ways, some of which may not be captured by area-level measures of risk.For example, differences in occupational exposure to hazards, the safety of available transport, experiences of domestic violence, and a range of individual differences in our day-to-day experiences all make quantifying the degree to which we can control mortality risk highly complex.

Limitations
One possible limitation of the study could be attributed to the level of abstraction of the measures of risk and health effort.By utilizing a general measure of health effort, the study does not capture potential relationships that could exist between perceived control over specific causes of death and those specific health behaviors that correspond most obviously with them.Previous research during the COVID-19 pandemic showed that higher levels of PUMR were associated with lower reported adherence to UK Government advice on diet, physical activity, and smoking (Brown et al., 2021a).Our analysis in this study did not investigate the effects of PUMR on specific health behaviors but did show that PUMR contributes to predicting overall health effort, above and beyond the effect of perceived of control over multiple causes of death.This is unsurprising, given that it is the level of general risk that one faces, which should affect motivation to invest in overall health.That is, it would be reasonable to predict that a measure of overall perceived control would affect a range of health behaviors, whereas perceptions related to a single hazard would be less likely to have such a generalized effect.Second, our measure of accuracy compared the perceived prevalence of different categories of avoidable death with UK mortality statistics and therefore did not reflect individual assessments of personal risk.Accuracy scores in this study are likely to have depended on an individual's knowledge of risk at a societal level, rather than an interpretation of the degree to which they are personally exposed to specific risks.Surprisingly, our measure of numeracy did not predict accuracy scores for the prevalence of risk.Previous research has reported that individual numeracy predicts the accuracy of risk estimations concerning breast cancer risk (Weil et al., 2015), general cancer risk (Vromans, 2022), and HIV (Ellis et al., 2014).It is possible that the lack of effect in this study is because we encountered a ceiling effect, with most participants scoring very highly for individual numeracy.Future studies may investigate the influence of numeracy on perceived risk by either using a more complex assessment of numerical understanding or possibly by limiting the time participants have to complete the survey item.We encountered a similar challenge with our measure of occupational exposure to risk.Most of the participants from our sample reported low levels of exposure to risk at work.The lack of variation in reported occupational exposure to risk may explain why this variable did not predict perceived control over risk due to injuries.Finally, though our measure of SDI demonstrated an "adequate" degree of internal consistency (Taber, 2018), future research may consider improving the SDI scale by refining and adding additional items to increase the reliability of this measure.

CONCLUSION
This study investigated the individual characteristics that predict levels of perceived control over mortality risk, and the degree to which perceptions influence self-reported health effort.Our findings show that overall perceptions of control over mortality risk may capture people's general sense of environmental risk, influenced by both exposures to hazards and the availability of resources to avoid threats to health and longevity.Conversely, individual assessments of control over specific risks may involve more deliberate, considered appraisals of risk.Our results show that someone's general sense of the uncontrollability of mortality risk in their environment may play a more notable role in determining individual investment in health behaviors than assessments of specific risk.Age and educational differences in the accuracy of risk estimations suggest that health interventions should target younger, less educated individuals to better align perceptions with objective levels of risk.Finally, given this study's additional support for the relationship between increased levels of PUMR and lower health effort, further investigation is needed into the extent to which PUMR accurately reflects objective measures of individual risk.

A C K N O W L E D G M E N T S
We received financial support from Prolific and the European Human Behaviour and Evolution Society for the development and delivery of the survey questionnaire associated with this study.

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors declare that there are no conflicts of interest.

F
I G U R E 1 Perceptions of control over causes of death by age.Note: Error bars represent 95% confidence intervals (n = 1463)."Younger" represents participants aged 45 or below, "Older" represents participants older than 45.Asterisks indicate a significant difference between younger and older participants in perceived control over the specific cause of death; "***" indicates p < 0.001.

F
Perceptions of control over causes of death by gender.Note: Error bars represent 95% confidence intervals (n = 1463).Asterisks indicate a significant difference between men and women (other genders not shown due to small numbers) in perceived control over the specific cause of death; "**" indicates p < 0.01, and "***" indicates p < 0.001.

F
I G U R E 3 Perceptions of control over causes of death by subjective discretionary income.Note: Error bars represent 95% confidence intervals (n = 1463)."LowSDI" represents participants who scored below the median value for subjective discretionary income (SDI), "HighSDI" represents those scoring the median value or above.Asterisks indicate a significant difference between LowSDI and HighSDI participants in perceived control over the specific cause of death; "**" indicates p < 0.01 and "***" indicates p < 0.001.

F
Prolific and the European Human Behaviour and Evolution SocietyD ATA AVA I L A B I L I T Y S TAT E M E N TAll data and code associated with this submission will be made available as part of the Open Science Framework (osf.io/dgwna).

TA B L E 1
List of study variables.
Overall effect of socioeconomic, demographic, and occupational exposure variables on perceptions of control over 10 causes of death.This table shows the results of a multivariate analysis of variance with respect to a multiple regression model comprising the predictors listed above and the following causes of death as outcome variables: cancers, circulatory disease, respiratory disease, alcohol and drugs, infectious disease, injuries, COVID-19, environmental risks, antimicrobial resistance, and catastrophic risks.
Comparing regression models predicting self-reported health effort.Model 1-Multiple linear regression model with socioeconomic and demographic variables predicting self-reported health effort.Model 2-Perceptions of control over 10 different causes of death are added to Model 1 as predictors.Model 3-Perceived uncontrollable mortality risk (PUMR) is added to Model 2 as a predictor.β represents standardized regression weights with 95% confidence intervals provided in square brackets below."**"indicates p < 0.01.RMSE represents the root mean squared error for each model.Abbreviation: SDI, subjective discretionary income. Note: