Silver Spoons in Toxic Soups: Predictors of Protective Action in Affluent Neighborhoods near a Contaminated Research Facility

ABSTRACT

The Santa Susana Field Laboratory (SSFL) in Ventura County, California, was among the first sites in the United States to conduct aerospace and nuclear research during the Cold War. Over the nearly 60 years it was in operation, several accidents at SSFL resulted in significant contamination of the surrounding environment. As a result, SSFL was designated as a Superfund eligible site in 1989, although not placed on the National Priorities List, and has since been the subject of numerous lawsuits and health investigations. Prior research near the SSFL has identified hazardous contaminants and possible associations with certain health outcomes, yet uncertainty surrounding the health effects of those exposures has persisted for decades. Despite this uncertainty, one characteristic of the SSFL distinguishes it from many other contaminated sites. The communities surrounding the SSFL are substantially wealthier than those typically studied near Superfund sites. This context allowed us to test protection motivation theory (PMT) as well as other predictors such as health anxiety, morbidity, and distance from the site in a population‐type that has been understudied in environmental risk research. Findings indicate that (1) autoimmune and endocrine/metabolic conditions were elevated among those living closest; (2) protective action was most strongly predicted by PMT variables, particularly threat appraisal, and (3) health anxiety partially mediated the relationship between morbidity and protective action.

1. Introduction

Initially built for Cold War aerospace and nuclear research, the Santa Susana Field Laboratory (SSFL) is now better known for its environmental legacy than its scientific achievements. Located in Ventura County, California, SSFL has been the focus of long‐running public health concerns in nearby communities (Reynolds et al. 1992; Morgenstern et al. 2007). In the United States, such sites fall within the scope of the U.S. Environmental Protection Agency's (EPA) Superfund program, which facilitates investigation and remediation (Ulibarri et al. 2020).

Operational from 1949 to 2006, activities at SSFL were undertaken by the Atomic Energy Commission (AEC), Boeing, the Department of Energy (DOE), and the National Aeronautics and Space Administration (NASA). Together, their research contributed to the United States’ first liquid‐fuel engine in 1950, satellite launch in 1958, and Moon landing in 1969 (NASA 2009). Although SSFL supported major experimental research, its activities produced a legacy of radioactive and chemical waste that contributed to widespread contamination of the air, groundwater, and soil on and off‐site (Zhao et al. 2005; Boice et al. 2006). The most notable incident was a partial nuclear meltdown of a sodium reactor in 1959 (Rucker 2009).

This accident, along with routine operational releases from the lab's day‐to‐day goings‐on led SSFL to be entered into EPA's Comprehensive Environmental Response, Compensation, and Liability Information System (CERCLIS) in 1980 (EPA 2007). Although SSFL was evaluated and deemed eligible for inclusion on the EPA's program for federal Superfund cleanups, it has not been added to the National Priorities List (NPL), and cleanup has therefore proceeded under state supervision (NASA 2009; Ulibarri et al. 2020).

Consistent with other such sites, concerns about potential health risks have influenced perceptions of the SSFL, despite mixed and inconclusive evidence regarding the health effects of site‐related exposures (Ulibarri et al. 2020). While living near such a site is not unique, SSFL is an instructive case because it is situated near some of the most affluent communities in the United States; people who are substantially wealthier than those typically studied near Superfund sites (Morgenstern et al. 2007; Trottier et al. 2025). According to a national study, 62% of households within 8 km of Superfund sites earn under $50,000 and 8% earn $100,000 or more (Trottier et al. 2025). In our sample, however, 15% of households within 8 km of SSFL earn under $50,000 and 56% earn $100,000 or more. Further, according to the most recent United States Federal Census, median household income within 24 km of SSFL is $104,047 and rises to over $200,000 within 1 km, compared with $86,587 in Los Angeles County and $107,667 in Ventura County¹. This setting raises an underexamined question for risk researchers: How do affluent people respond to risks posed by a contaminated site?

Thus, we apply protection motivation theory (PMT) to predict protective action in communities living within 24 km of SSFL. Prior studies have examined PMT's original mechanisms (e.g., perceived severity, perceived vulnerability, response efficacy, self‐efficacy, and response costs) and tend to model them as mediators between threat information and protective action (Rogers 1975). Yet, relatively few have studied PMT beyond these core appraisals. Recent reviews therefore recommend integrating additional constructs to enhance PMT's predictive power and improve its generalizability (Siegrist and Árvai 2020; Balla and Hagger 2025). Accordingly, we incorporate measures of health anxiety, morbidity, and distance into our PMT analysis of residents living near SSFL. Therefore, our contribution is to test PMT's applicability in this context, identify any theoretical deviations, and to extend PMT to include measures regarding health and distance. We believe it is important to examine this theory in such an unusual context as SSFL not only to provide pertinent information to the public but also to inform risk researchers and policymakers of any limitations within a framework on which we so often rely.

1.1. Protection Motivation Theory and Health Anxiety

PMT is a social‐cognitive model used to understand and predict how people respond to health risks (Rogers 1975). It includes two main processes. The first is threat appraisal, which refers to how people perceive risks based on how serious they think it is and how vulnerable they feel to it (i.e., perceived severity and perceived vulnerability). The second is coping appraisal, which involves evaluating how effective protective actions are and whether people feel capable of taking those actions (i.e., response efficacy and self‐efficacy). A measure of perceived risk results from these mechanisms, which may move people to take protective action. Most literature finds that people are more likely to take protective action when they see a risk as severe, feel vulnerable to it, believe that protective actions will be effective, and have confidence in their ability to take those actions (e.g., van der Pligt 1998). However, motivation may decrease if people see immediate benefits in avoiding protective action, or if the costs of acting (e.g., in terms of effort, money, or time) are too high (Norman et al. 2005).

Along with PMT variables, we include a clinical measure of health anxiety, which is defined as heightened vigilance and concern about health threats (Asmundson and Taylor 2020). We suspect that health anxiety may be especially relevant at SSFL where contamination‐related information is uncertain, government guidance is limited, and large‐scale remediation remains incomplete (EPA 2007). It is possible that health anxiety may explain additional variance in protective actions within PMT and potentially operate beyond risk perceptions (Olatunji et al. 2011; Parkhill et al. 2011). There is a chance, however, that severe health anxiety could lead to resignation or avoidant behaviors (Rachman 2012; Kikas et al. 2024) and thus result in less protective action. We include health anxiety in our analyses, among other variables such as morbidity and distance, to explore its effects on engagement in protective actions.

1.2. Morbidity and Distance

Epidemiological studies have produced mixed findings regarding health risks near the SSFL. Although early investigations found no consistent increase in overall cancer incidence (Wright and Perkins 1990; Reynolds et al. 1992), later research reported higher incidence rates of thyroid, bladder, lymphopoietic (blood and lymph tissue), and upper aerodigestive cancers among residents living within 3 km of the site than among those living farther away, suggesting a possible association between residential proximity and cancer incidence while stopping short of establishing a causal relationship (Morgenstern et al. 2007).

The spread of contaminants at SSFL has been exacerbated by drought, Santa Ana winds, wildfires, and the site's elevation. For example, the most recent major fire affecting SSFL was the 2018 Woolsey Fire (Kaltofen et al. 2021). It began at SSFL, burned 80% of the site, and spread across 390 square km in Los Angeles and Ventura counties.

Events like these can increase the risk of developing chronic health conditions such as cancer, cardiovascular disease, neurological disorders, and respiratory illness due to exposures to heavy metals, toxic chemicals, and fine particulate matter from Superfund sites (Landrigan et al. 2015; Ulibarri et al. 2020). Indeed, fine particulate matter and other contaminants at Superfund sites have well‐established relationships to diseases beyond cancer, particularly cardiovascular and respiratory conditions. Particulate exposure, especially, can induce systemic inflammation and oxidative stress, and thus increase the risks of heart and lung disease (Stephens et al. 2022; Długosz‐Lisiecka et al. 2025).

Despite evidence of contamination extending up to 24 km from SSFL, partly due to the 2018 Woolsey Fire (Kaltofen et al. 2021), studies have yet to examine health outcomes beyond an 8 km radius (Morgenstern et al. 2007). Thus, in this study, we examine a broad range of chronic health diagnoses within 24 km of the site (Figure 1).

FIGURE 1.

SSFL and surrounding areas.

1.3. Research Questions

We explore the following research questions.

1. What is the distribution of health diagnoses among individuals living near SSFL?

2. To what extent do (a) traditional PMT variables, (b) health anxiety, (c) morbidity, and (d) distance predict protective action?

3. Does health anxiety mediate the relationship between morbidity and protective action?

2. Methods

2.1. Sample

Participants were recruited by the Understanding America Study (UAS) using a stratified sampling approach for respondents living within 24 km of the SSFL. UAS invited 212 UAS households and 2,483 non‐UAS households to complete our survey. All invitees received four mailings spaced 2 weeks apart. In line with prior research on reciprocity in survey participation (Becker 2023), the first mailing included a pre‐notification postcard with a $20 Amazon gift card as an incentive. Subsequent mailings included a priority mailer, a reminder letter, and a final reminder postcard. A total of 54 respondents from the UAS database completed the survey (26% response rate). Of the 2,483 invitations mailed to non‐UAS households, 44 were returned as undeliverable; of the remaining 2,439 eligible respondents, 421 completed the survey (17% response rate). Thus, the final sample size was 475 participants (18% total response rate across groups). Human subjects research approval was obtained from the UAS, which oversees research projects conducted by the University of Southern California.

2.2. Design

All survey materials were presented in English and Spanish and all participants provided informed consent prior to participation. The survey was administered online through the UAS platform. All questions in the survey were included and in the same order for all participants. The first question of the survey asked all participants if they had heard of SSFL. When participants reported that they had, they proceeded to take the survey and responded to questions about perceived vulnerability, self‐efficacy, protective actions, response efficacy, response costs, risk perceptions, health anxiety, morbidity, and demographics.

2.3. Measures

2.3.1. Protection Motivation Theory Variables

Threat appraisal was estimated using two measures: vulnerability and severity. Vulnerability was measured using a modified version of the climate change distress and impairment scale (Hepp et al. 2023). We adapted the scale into six items and modified the questions to fit the contamination context (α = 0.85). For example, “When I think about climate change, I worry about the future” became “When I think about the contamination, I worry about what living in my home will mean for me in the future.” Responses were measured on a scale from strongly disagree (1) to strongly agree (5). For severity, participants were asked to report how severe they perceived the contamination from SSFL on a scale of 0–100. Because the vulnerability and severity items were measured on different scales, each was first standardized using z‐scores before being averaged to create a composite threat appraisal variable.

Coping appraisal was estimated using two measures: response and self‐efficacy. For response efficacy, participants were asked to rate how effective they believed each protective action to be in mitigating exposure to contamination (whether they engaged in the action or not) on a scale from (1) not at all effective to extremely effective (5) (α = 0.93). Self‐efficacy was measured using the generalized self‐efficacy scale, which measures how competent people perceive themselves to be when managing difficult situations (Schwarzer and Jerusalem 1995; Warner and Schwarzer 2020). Items were measured on a 4‐point scale from not at all true (1) to exactly true (4). We adapted the scale to include five items and modified the questions to fit the contamination context (α = 0.85). For example, “I am confident that I could deal efficiently with unexpected events,” was modified to “I am confident that I could efficiently handle unexpected events caused by the contamination.” We measured responses on a scale from strongly disagree (1) to strongly agree (5). Next, the self‐efficacy and response efficacy items were averaged to create a composite efficacy variable. We elected to analyze response costs separately to isolate the role of perceived financial burden in predicting protective action. Participants rated the anticipated financial cost of reducing their SSFL‐related risk using a scale from (1) no financial cost to extreme financial cost (5).

To assess risk perceptions, participants were asked to report how risky they perceived the contamination from SSFL on a 0–100 scale for their personal health (α = 0.91). And to assess protective action, participants were asked to report on what activities, if any, they engaged in to avoid contamination from SSFL. These activities included (1) not eating fruits, vegetables, and/or herbs grown in participants’ gardens, (2) not eating produce from farms near SSFL, (3) not hiking on trails near SSFL, (4) leaving shoes outside of the home to avoid bringing in dust and dirt, (5) filtering tap water or using bottled water for drinking and cooking, (5) avoiding time outside in their yards, especially during rain and wind, (6) using an air purifier in their home, (7) regularly testing soil for chemicals and contaminants, and (8) “other” actions which were captured as an open‐ended variable (open‐ended responses were too few in number to analyze and report in this study). Participant responses were measured on a binary scale with “yes” indicating that an action was taken, and “no” indicating that an action was not taken; and the number of actions taken were then summed.

2.3.2. Health Anxiety

Health anxiety was measured using the short health anxiety inventory (α = 0.86), which is a measurement widely applied across academic and clinical settings (Abramowitz et al. 2007). The assessment included 14 items that measure elements of health anxiety like thought intrusion, bodily preoccupation, and fear of illness. Items were measured on a 4‐point scale, scored 0–3. Participants answered items to receive an aggregate score from 0 to 42. Those who scored 0–27 represent no or mild levels of health anxiety, 28–32 represent moderate health anxiety, and 33–42 represent substantial health anxiety.

2.3.3. Morbidity

Participants were asked to report any diagnoses of chronic health disorders they received since living in their current home. Conditions such as autoimmune disorders, cancer, cardiovascular disorders, metabolic or endocrine disorders, neurological disorders, and respiratory disorders were included due to the established association of these conditions with Superfund site exposure (Stephens et al. 2022; Długosz‐Lisiecka et al. 2025). Morbidity was calculated as the total number of diagnoses reported by each participant.

2.3.4. Distance, Income, and Demographic Variables

A measure of physical distance was embedded in the sampling strategy. UAS used residential addresses to create 1.6‐km (1 mile) distance bands from SSFL. For descriptive reports (such as in Table 1 and Table S1), distance from SSFL was categorized into three bands (0–8 km, 9–16 km, and 17–24 km). These bands were defined based on prior health studies (Kaltofen et al. 2021; Morgenstern et al. 2007) and contamination reports and represent equal intervals. For all other analyses, distance was treated as a continuous variable (0–24 km).

TABLE 1.

Morbidity by strata and diagnosis (N = 475).

Morbidity by strata	All (N = 475)	0–8 km (n = 296)	9–16 km (n = 129)	17–24 km (n = 50)	p
% with one disclosed diagnosis	22%	22%	24%	16%	—
% with two disclosed diagnoses	11%	13%	7%	4%	—
% with three disclosed diagnoses	3%	4%	3%	0%	—
% with four or more disclosed diagnoses	1%	1%	1%	0%	—
Diagnosis type by strata
% with an autoimmune disorder	16%	18%	16%	4%	<0.05
% with cancer	11%	11%	10%	8%	0.78
% with cardiovascular disorder	5%	7%	3%	2%	0.16
% with a metabolic/endocrine disorder	12%	15%	9%	2%	0.02
% with a neurological disorder	5%	6%	2%	4%	0.29
% with a respiratory condition	8%	9%	9%	8%	0.48
% with “other” diagnosis	11%	11%	10%	14%	0.75
% with an undisclosed diagnosis	12%	11%	15%	8%	0.42
% who do not report a diagnosis by strata
	63%	59%	65%	80%	0.04

p‐values are based on Pearson chi‐square tests comparing diagnosis prevalence across distance strata. Participants were able to select more than one diagnosis.

Income was measured by asking participants to disclose their annual household earnings (without specifying gross or net income or household size). The options from which they could choose were categorized into 16 ranges, capped at $150,000 or more. Data for age, duration of residence, education, ethnicity, and sex were also collected.

2.4. Analyses

To address our first research question, we computed descriptive statistics and conducted a robust regression analysis to examine differences in morbidity as a function of distance to SSFL, while controlling for age, duration of residence, and income. To address our second research question, we estimated a multiple linear regression model predicting protective actions from standard PMT variables, health anxiety, morbidity, distance, and income. To address our third research question, we estimated a mediation model using PROCESS Model 4 (Hayes 2022) to examine whether health anxiety mediated the relationship between morbidity and protective actions.

3. Results

3.1. Demographics and Correlations

Of our 475 participants, 74% reported being 45 years or older, 42% identified as male, 80% identified as white, 62% lived fewer than 8 km from the SSFL, and 37% reported at least one chronic health diagnosis. The income distribution was concentrated in the upper categories, with more than half of respondents reporting household incomes of $100,000 or more. Additional demographic characteristics can be found in the supplemental materials (Table S1).

A Pearson's correlation indicated that among PMT variables, threat appraisal had a strong and positive correlation with risk perception and a moderately positive correlation with response costs and protective actions. Risk perception and response costs also had a moderately positive correlation with protective actions. Efficacy was not correlated with protective actions, threat appraisal, or risk perception; and had a weak negative correlation with response costs. Secondly, among health variables, health anxiety was positively correlated with threat appraisal, risk perception, response costs, protective actions, and morbidity; and morbidity was positively correlated with threat appraisal, risk perception, and protective actions. Thirdly, distance and income had generally weak correlations with other variables (Table 2).

TABLE 2.

Pearson's correlation of all participants (N = 475).

	M	SD	1	2	3	4	5	6	7	8	9	10	11
Threat appraisal	0.01	0.79	1.00
Efficacy	2.57	0.56	−0.04	1.00
Response costs	2.83	1.28	0.43 **	−0.15 **	1.00
Risk perception	50.36	30.15	0.78 **	0.02	0.47 **	1.00
Protective actions	1.31	1.53	0.51 **	0.09	0.35 **	0.49 **	1.00
Health anxiety	12.24	5.85	0.38 **	−0.06	0.27 **	0.45 **	0.33 **	1.00
Morbidity	0.57	0.88	0.22 **	−0.11 *	0.18 **	0.24 **	0.28 **	0.30 **	1.00
Distance	5.23	3.19	−0.04	0.18 **	−0.08	−0.09	−0.08	−0.02	−0.13 **	1.00
Income	13.58	3.47	−0.02	−0.09	−0.03 *	−0.11 *	0.00	−0.14 **	−0.09 *	−0.06	1.00
Education	12.40	2.00	−0.09	−0.08	−0.04 **	−0.14 **	−0.04	−0.07	−0.03	0.04	0.35 **	1.00
Age	55.29	15.70	−0.08	−0.12 **	−0.08	−0.08	−0.12 **	−0.13 **	0.19 **	−0.10 *	−0.02	0.00	1.00

Note. ^* p < 0.05; ^** p < 0.01.

3.2. Those Living Closest to SSFL Report Higher Morbidity

Reports of diagnoses generally decreased with distance from the SSFL (Table 1). Chi‐square analyses indicated significant differences across distance strata for autoimmune disorders and metabolic/endocrine disorders whereas differences for cancer, cardiovascular, neurological, respiratory, and undisclosed diagnoses were not statistically significant. The 0–8 km and 9–16 km strata exhibited similar levels of morbidity overall, whereas participants living 17–24 km from the SSFL reported fewer diagnoses. A multiple regression model examined total morbidity as a function of distance from the SSFL, duration of residence, age, and income and was statistically significant but explained a modest proportion of variance (Table 3).

TABLE 3.

Robust regression analysis on morbidity by distance to SSFL, duration of residence, age, and income (N = 475).

Predictor	B	SE (HC3)	t	p	95% CI	ηp 2
Intercept	0.51	0.25	2.08	0.04	[0.03, 1.00]	0.01
Distance	−0.03	0.01	−2.91	0.00	[−0.05, ‐0.01]	0.02
Duration of residence	0.03	0.02	1.62	0.11	[−0.01, 0.06]	0.01
Age	0.01	0.00	2.57	0.01	[0.002, 0.01]	0.01
Income	−0.02	0.01	−1.77	0.08	[−0.05, 0.002]	0.01

Note. F(4, 470) = 7.90, p < 0.001, ηp² = 0.06. The data violated the assumption of normality and homogeneity, so we corrected for these violations by using the HC3 robust standard error and 5000 bootstrapped samples (Hayes and Cai 2007).

3.3. PMT and Health Variables Predict Protective Action

PMT variables predicted protective actions. Threat appraisal, efficacy, and response costs were significant positive predictors of protective action, while risk perception was not significant (Table 4). Higher health anxiety and higher morbidity also predicted more protective actions, and older age predicted fewer protective actions, though the effect of age was small. Regarding effect size, threat appraisal and morbidity were moderate; efficacy and response costs were small to moderate; and health anxiety and age were small. Risk perception, distance, income, and education were not significant predictors in this model.

TABLE 4.

Linear regression on protective actions (N = 475).

Predictor	B	SE	β	t	p	95% CI
Threat appraisal	0.59	0.12	0.31	5.04	<0.001	[0.36, 0.83]
Efficacy	0.41	0.11	0.15	3.80	<0.001	[0.20, 0.62]
Response costs	0.15	0.05	0.12	2.79	0.01	[0.04, 0.25]
Risk perception	0.01	0.00	0.10	1.58	0.12	[−0.00, 0.01]
Health anxiety	0.02	0.01	0.09	2.09	0.04	[0.00, 0.05]
Morbidity	0.29	0.07	0.17	4.06	<0.001	[0.15, 0.43]
Distance	−0.03	0.02	−0.06	−1.56	0.12	[−0.07, 0.01]
Income	0.02	0.02	0.05	1.27	0.20	[−0.01, 0.06]
Education	0.01	0.03	0.02	0.43	0.67	[−0.05, 0.07]
Age	−0.01	0.00	−0.09	−2.28	0.02	[−0.02, −0.00]

R² = 0.35, adjusted R² = 0.34, F(10, 464) = 25.31, p < 0.001.

3.4. Health Anxiety Mediates the Relationship Between Morbidity and Protective Actions

We estimated mediation using PROCESS Model 4. Morbidity was positively associated with health anxiety, and health anxiety was positively associated with protective actions. The indirect effect was small but significant, and the direct effect of morbidity on protective actions remained significant, which indicates partial mediation (Figure 2, and see Table S3 in the supplemental materials).

FIGURE 2.

Mediation of the relationship between morbidity and protective actions by health anxiety. Note: Values shown are standardized regression coefficients (β). Unstandardized coefficients and standard errors are reported in Table S3.

4. Discussion

4.1. Morbidity and Distance

Previous health research at SSFL focused primarily on cancer incidence among communities living within 8 km of the site (Morgenstern et al. 2007). But cancer is only one of several illnesses reported near Superfund sites (Landrigan et al. 2015), and SSFL‐related contamination has been documented up to 24 km from the site (Kaltofen et al. 2021). For these reasons, the current risk study looked at a broader range of chronic health diagnoses across a larger area near SSFL.

Consistent with findings at other sites (Landrigan et al. 2015; Dunnick et al. 2025), participants living closest to SSFL reported the highest number of health diagnoses in several health categories. Autoimmune disorders and endocrine/metabolic disorders were significantly more prevalent among participants living closer to the SSFL, whereas cancer prevalence did not differ significantly across distance strata. Overall, participants living within 0–8 km and 9–16 km of the SSFL reported similar levels of morbidity, whereas those living 17–24 km from the site reported fewer diagnoses.

Nationally, an estimated 129 million people in the United States (about 38%) have had at least one major chronic health diagnosis (Benavidez et al. 2024b). The prevalence of chronic health issues, however, tends to be much lower in affluent communities than in poorer communities (Benavidez et al. 2024a). Given our sample's relatively high median household income, one might expect morbidity to fall below national levels, were it not for the presumed risk of contamination near SSFL. Instead, 37% of participants in our study reported at least one chronic health diagnosis.

4.2. Protection Motivation Theory and Risk

Consistent with prior research (e.g., Balla and Hagger 2025), PMT variables in this study were found to be robust predictors of protective action in the communities surrounding SSFL. Our findings do, however, reveal deviations from standard PMT results that merit discussion. First, while prior studies have found that coping ability and perceived efficacy are more predictive of protective action than threat appraisal (Milne et al. 2000; Balla and Hagger 2025), threat appraisal emerged as the strongest PMT predictor of protective action in our model. Efficacy only had a small significant effect (results for the traditional coping appraisal variable are reported in Table S2 in the supplemental materials). Secondly, response costs positively predicted protective action. Thirdly, risk perceptions did not significantly predict protective action. Lastly, health anxiety and morbidity predicted protective action, though distance from the SSFL did not.

First, the few risk studies that include an analysis of income state that those with higher incomes have diminished feelings of threat due to their greater and known capacity to cope with costs and their stronger sense of efficacy (Lo 2014). Consistent with these findings, income was negatively associated with both health anxiety and risk perception in our sample. Yet, income was not significantly associated with threat appraisal. One explanation is that threat appraisal was a stronger predictor of protective action than efficacy because residents are repeatedly reminded of potential SSFL‐related harms, yet have few clear, effective ways to reduce their exposure regardless of cost. Environmental contamination is also largely invisible, so residents cannot readily confirm whether they are being exposed or which steps, short of moving, reliably reduce potential health impacts. Uncertainty paired with regular SSFL‐related media attention, reoccurring wildfire events, and community activism with local groups may keep the threat of the SSFL and its contamination top of mind.

Indeed, concern about a threat is more likely to motivate action when it is salient (Horner et al. 2023). Practically, though, there are few high‐impact protective actions available. Unlike fires or flooding, where residents can see the risks and invest in resilient building materials or property‐level buffers (Nagisetty et al. 2020), most of the SSFL‐related precautions involve limiting exposure to the outdoor environment, which may be challenging or seem arbitrary to people whose homes and daily routines are near the site. The risk perception literature also makes clear that self‐efficacy is a phenomenon that develops through mastery, physiological, and affective states (Waddington 2023). Thus, efficacy beliefs may be difficult to sustain without opportunities to act and to receive visible reinforcement. If residents are also coping with illness, distress, or frustration over delayed cleanup, efficacy beliefs are likely to diminish over time regardless of orientation. Even so, efficacy may be a weaker predictor of protective action if residents continue to take precautions because the threat remains salient and the risks are believed to be unmitigated.

Second, we found that higher perceived response costs predicted more protective action, which runs counter to the usual expectation that costs discourage behavior (Rogers 1975; Milne et al. 2000). This finding may reflect how response costs were experienced and interpreted in our study. For example, many of the behaviors we measured were low‐cost avoidance behaviors (e.g., not eating fruits, vegetables, and/or herbs grown at home) rather than costly actions and our response cost item focused on the perceived financial cost of lowering one's risk overall. Residents who were already taking precautions, especially those involving purchases such as water filtration or soil testing, may therefore have been more aware of the costs and tradeoffs involved in risk reduction.

In this sense, reported “costs” may indicate engagement with protective action rather than functioning as a deterrent. Of course, this finding may also reflect the socioeconomic context of our sample. In a relatively affluent community, financial and practical burdens may be less prohibitive for residents who want to act (Bindley et al. 2023; Lusk et al. 2023), which could allow protective behavior to remain high even when risk reduction is perceived as costly. Given the setting, there is a likelihood that cost is interpreted as a signal of effectiveness for residents, since higher prices often cue perceptions of greater quality and efficacy (Díaz‐Lago et al. 2023; Schiessl 2023).

Third, a generalized measure of risk perception did not predict protective action once threat appraisal and efficacy were included. This result likely reflects overlap with threat appraisal, as the two were strongly correlated (Table S2). Similar null effects for broad risk measures have also been reported elsewhere, especially when more specific, action‐focused appraisals are included (Slovic et al. 2004). Here, “risk” may have been hard for residents to judge in a concrete way because potential exposure is invisible, and like all risks, somewhat uncertain. As a result, respondents may have answered the risk question based on general worry rather than on a clear personal estimate, which threat appraisal already captures more directly as vulnerability and perceived severity (Table S2).

Fourth, health anxiety and morbidity both predicted protective actions, although their effects were small. One interpretation is that health anxiety heightens sensitivity to uncertainty and vulnerability regarding health, so it may amplify precautionary, “just in case” responses in much the same way as threat appraisal. Likewise, we expect that morbidity may have operated less as a motivational factor and more as a routine, since those managing chronic conditions may be more attentive to symptoms during smoke, dust, or stress and more likely to follow ongoing medical guidance that overlaps with the protective behaviors we assessed.

4.3. Protective Actions and Health Anxiety as a Mediator

We examined how much health anxiety accounts for variation in protective action. The model tested whether health anxiety mediated the relationship between morbidity and protective action and the model fit the data. Total morbidity was associated with higher health anxiety; and health anxiety predicted more protective action (Figure 2, Table S3). However, morbidity remained a significant predictor after accounting for health anxiety, so the mediation was only partial. This finding aligns with an earlier observation, where people with more health problems may take more precautions both because they feel vulnerable and because they are already monitoring symptoms and managing their health in ways that overlap with the protective actions measured.

4.4. Limitations and Future Research

Several research limitations should be noted. First, protective actions were measured only as ever taken, not their frequency. Second, morbidity relied on self‐reported diagnoses without confirmation by a medical professional or information on timing and recurrence. Had we captured the frequency and recurrence of individual diagnoses, rather than only whether a respondent had ever been diagnosed with a condition since living in the area, the differences across groups may have been more pronounced. In other words, our measure allowed respondents to indicate each diagnosis category only once (e.g., “cancer”), without capturing multiple occurrences, or different types of diagnoses within each category over time. As a result, individuals, particularly those living closest to SSFL who often had longer residence durations, may have experienced repeated or multiple diagnoses that were not fully reflected in our broad, single‐instance measures. Further, because morbidity was measured as prevalence (number of diagnoses) rather than incidence (diagnoses over time), the associations we report may partly reflect who is still living in the area at the time of the survey. Residents who became ill may have moved away or died, and others may have moved in or remained after SSFL risks became more widely known; both processes can influence the observed prevalence. Third, standard measures of household income were used in this study, meaning that that highest income range measured was “$150,000 or more.” Future research in high‐income or high‐net‐worth communities should increase the upper bound of the income measure and consider additional measures of wealth.

Interpretations of our data should also consider the survey design. For example, participation may have been more likely among those with greater concern about SSFL, or among those with relevant health experiences. Finally, relocation may be an important protective response to risks from SSFL. However, we had no effective means of measuring this variable. If the most concerned residents had moved away, our sample may underrepresent those who perceive the highest level of risk.

5. Conclusion

This study addressed three questions. First, we documented which health diagnoses were prevalent among residents living near the SSFL and summarized overall morbidity in the sample. Second, we tested predictors of protective action. Protective action was most strongly associated with threat appraisal, while efficacy was less predictive than standard PMT results would suggest. Third, we evaluated whether health anxiety helps explain why some residents take more action than others. We found that health anxiety partially mediated the relationship between total morbidity and protective action.

Practically, our findings may reflect the broader social and physical area surrounding the SSFL. The site operated for decades, sits high in the Simi hills and largely out of view, and is connected to institutions such as NASA that many Americans view favorably². In this setting, residents may not readily attribute health risks to SSFL, regardless of how close they live or how many health problems they report. The surrounding area also includes high‐income neighborhoods such as Bell Canyon (median home price >$2.5 million) and Hidden Hills (median home price >$6 million) with celebrities such as Kim Kardashian, Kourtney Kardashian, Joe Rogan, Miley Cyrus, Gerard Butler, and others. It is therefore plausible that some residents tolerate more risk when the location offers prestige, strong real‐estate value, and a sense of stability or control (Diener et al. 1985; Baumeister et al. 2003).

These considerations suggest two directions for future work by risk analysts. First, though it is understandable that physical distance is the most appropriate measure for modeling exposure to toxins, perceived distance may be a better—or, at the very least, additional—measure for modeling behavioral responses to risks (Liberman and Trope 2014; Tanner and Árvai 2018). Second, in a wealthy and potentially influential community, protective action may be more observable at the institutional level (e.g., advocacy, litigation, testimony at hearings) than the individual level, especially since protection near SSFL depends largely on decisions by government and responsible parties.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supplementary Table S1 Demographic characteristics of participants (N = 475).

Supplementary Table S2 Linear regression on protective actions including traditional coping appraisal (N = 475).

Supplementary Table S3 Mediation of the Relation Between Morbidity and Protective Actions by Health Anxiety (PROCESS Model 4, N = 475)

Supplementary Table S4 Multiple regression predicting protective actions with income as moderator (N = 475).

Data Availability Statement

Understanding America Study, Survey 661.