How are mental representations of asthma triggers and symptoms related to interpersonal risk perceptions? A psychometric investigation of caregivers of children with asthma

Abstract

Objective:

Parents and guardians (hereafter caregivers) make decisions for their children’s medical care. However, many caregivers of children with asthma struggle to understand their child’s illness. We used the psychometric paradigm to investigate how caregivers conceptualize, or mentally represent, asthma triggers and symptoms and how these representations are linked to perceived asthma exacerbation risk.

Methods:

We asked 377 caregivers of children with asthma across the U.S. to rate 20 triggers or 20 symptoms along 15 characteristics. Caregivers also indicated their perceived risk of their child having an asthma exacerbation (hereafter interpersonal risk perceptions). Using principal components analysis, we extracted key dimensions underlying caregivers’ ratings on the characteristics. Then we related the triggers’ and symptoms’ scores on the dimensions to caregivers’ interpersonal risk perceptions.

Results:

Interpersonal risk perceptions were higher for triggers with high ratings for the dimensions severe and relevant, and negative affect—yet manageable, but not chronic—yet unpredictable. Risk perceptions were also higher for symptoms with high ratings for the dimensions severe and unpredictable, and relevant and common, but not self-blame or manageable despite unknown cause.

Conclusion:

By identifying key dimensions underlying caregivers’ mental representations of asthma triggers and symptoms, these findings can inform a new approach to asthma education.

Introduction

Decades of research have demonstrated that adults’ perceptions of the riskiness of potential health hazards depend greatly on their mental representations of the hazard (Bronfman et al., 2008; Fischhoff et al., 1978; Slovic, 1987; Waters et al., 2017). However, little research has examined how the cognitions and affects that structure people’s conceptualizations and understanding of a hazard (i.e., mental representations) shape adults’ perceptions of their child’s risk of experiencing a negative health outcome (i.e., interpersonal risk perceptions). This is an important oversight; because parents and guardians (hereafter caregivers) often make medical decisions on behalf of their child (American Medical Association (AMA), 2023; Klok et al., 2011), identifying their mental representations and the relationship between their mental representations and interpersonal risk perceptions could be important for developing interventions for improving the health outcomes of children with complex chronic health conditions.

Our goal in this article is to identify caregivers’ mental representation of asthma triggers and symptoms and to relate key dimensions underlying these representations to caregivers’ interpersonal perceived risk of their child having an asthma exacerbation. To that end, we rely on the psychometric paradigm, which has been pioneered to identify key dimensions underlying laypeople’s risk perceptions (Fischhoff et al., 1978). We thus aim to demonstrate the utility of the psychometric paradigm for exploring caregivers’ mental representations of other complex chronic health conditions in children (e.g., diabetes, cystic fibrosis, sickle cell, juvenile rheumatoid arthritis, cancer, depression).

Biomedical, Clinical, and Social Contexts of Pediatric Asthma

Pediatric asthma is a chronic health condition in which a child’s airways are inflamed (National Heart Lung and Blood Institute (NHLBI), 2022). The airways constrict when the child encounters a substance that irritates them, and this constriction limits oxygenation. Health care providers often advise caregivers to help their child avoid substances that might trigger their child’s asthma, such as mold in the home, respiratory viruses, and allergens. They also provide caregivers information about common symptoms, such as wheezing and coughing. Most asthma is managed at home with inhaled corticosteroids to reduce chronic lung inflammation (often referred to as a controller, preventer, or maintenance medicine) and inhaled bronchodilators to alleviate acute breathing problems and airway constriction (referred to as rescue or quick relief medications). However, sometimes breathing problems become so bad that the rescue inhaler fails to completely alleviate symptoms and the child must receive more intense care (e.g., a short course of oral prednisone or a visit to the emergency department). In these cases, the child is said to have had an exacerbation (Fuhlbrigge et al., 2012).

Pediatric asthma is the most common childhood chronic health condition globally (Asher & Pearce, 2014), and is the most common pediatric lung disease in the US (Zahran et al., 2018). Approximately 4.2 million children under the age of 18 in the U.S. have asthma (Centers for Disease Control (CDC), 2022), and over 640,000 children per year visit U.S. emergency departments to treat asthma exacerbations (Centers for Disease Control (CDC), 2018). Asthma incidence and exacerbations are higher among children who are members of marginalized communities, including people who identify as Black, multiracial, Puerto Rican, or American Indian, Alaska Native, or Native Hawaiian (Asthma and Allergy Foundation of America, 2020; Centers for Disease Control (CDC), 2022). Children whose caregivers have low incomes also experience disproportionate rates of asthma. Asthma disparities originate in large part from national, local, and corporate policies that result in marginalized communities having increased exposure to asthma triggers and more limited access to health care services that help prevent and manage symptoms (Asthma and Allergy Foundation of America, 2020; Harris, 2019; Hunleth et al., 2020; Ruiz et al., 2022).

Caregiver Understandings of Asthma, Triggers, and Symptoms

Many factors at multiple levels of influence can make it difficult for caregivers to understand their child’s asthma and to determine their child’s risk of having an exacerbation. These factors include health care policies that limit access to pediatric pulmonologists and asthma educators (Association of American Medical Colleges (AAMC), 2020), person-level biological differences that produce wide variability in sensitivity to triggers (Asthma and Allergy Foundation of America, 2020), and the nature of asthma itself, in which ambiguous symptoms and bodily sensations complicate symptom interpretation (e.g., tightness in the chest could indicate an impending exacerbation or, more innocuously, anxiety; breathlessness from exertion could be misinterpreted as an asthma symptom) (Peterson et al., 2002; Spray et al., 2022). In addition, uncertainties about when and why asthma symptoms will occur can produce anxiety in caregivers, even in the absence of overt symptoms (Spray et al., 2022). Consequently, caregivers generate their own understandings of asthma. These understandings are produced via a complex process of collecting, integrating, and synthesizing information from multiple sources, including health care providers (when accessible and trusted), embodied sensations, trial and error experience, observation of co-occurring events, comparisons with other children with asthma, and advice from family members who have experience with asthma (Arcoleo et al., 2015; Heyduck et al., 2015; Hunleth et al., 2020; Ruiz et al., 2022; Shepperd et al., 2022; Spray et al., 2022). Caregivers then use these understandings to assess their child’s risk in different situations and to make decisions about courses of action (e.g., send an extra quick relief inhaler with the child when they are away from the caregiver, restrict the child’s engagement in physical activity, or avoid a particular family member’s house) (Ruiz et al., 2022; Spray et al., 2021; Spray et al., 2022).

Importantly, despite these Herculean efforts, many caregivers’ understanding of asthma as a health condition differs from the understanding of asthma that are accepted among the biomedical research and clinical practice communities (i.e., the biomedical model; see also the subsection above). For example, although some caregivers endorse the view that asthma is a chronic health condition, many view asthma as an acute and episodic condition that is present only when symptoms occur (Bokhour et al., 2008; Halm et al., 2006; Heyduck et al., 2015; Spray et al., 2022). Such inconsistencies likely occur because chronic lung inflammation is not directly observable by caregivers; rather caregivers can only detect acute symptoms such as audible wheezing and facial discoloration. Thus, a caregiver who relies on direct observation of symptoms would understandably conclude that their child only “has asthma” when he or she has symptoms.

Caregivers’ understanding of asthma triggers and symptoms are particularly important; without adequate knowledge of what triggers asthma-related breathing problems in their child and what early symptoms look and sound like, caregivers are less likely to determine that their child is at risk of an impending exacerbation and cannot avoid or address the problems before they progress in severity. Caregivers’ understanding of triggers and symptoms can involve elements that are consistent with the biomedical model, and also elements that are inconsistent with the biomedical model (Peterson-Sweeney et al., 2003). For example, in one study caregivers indicated that mold is a key asthma trigger (consistent with biomedical and clinical perspectives), but few caregivers identified house pests as potential triggers (inconsistent with the biomedical model) (Arcoleo et al., 2015). Complicating matters further is that the cause of many symptoms is ambiguous; a cough could indicate an impending asthma exacerbation needing immediate treatment, or it could indicate the beginning of a simple cold that will pass without incident (Spray et al., 2022). Of course, it is possible that a child may encounter an asthma trigger or display an asthma symptom that raises their risk of exacerbation yet is unrecognized by the biomedical model. However, the literature suggests that caregivers whose understanding of triggers and symptoms is more consistent with the biomedical model may be more likely to have children with better controlled asthma (Achstetter et al., 2019). Nevertheless, ambiguities and uncertainties about asthma triggers and symptoms can make understanding asthma—and asthma management decisions—more difficult (Spray et al., 2022).

Several interventions attempted to bring caregivers’ understanding of asthma, triggers, and symptoms in line with the biomedical model. These interventions were mostly developed either atheoretically or based on the Common Sense Model of Self-Regulation (Leventhal et al., 2003). In brief, the Common-Sense Model (CSM) asserts that patients’ (and caregivers’) understanding of illness is shaped, in part, by their beliefs about its identity (e.g., illness name and symptoms), timeline (e.g., duration, chronicity or acuity), consequences (e.g., severity), cause (e.g., genetics, environmental exposure), and controllability (e.g., treatable, preventable). Interventions based on the CSM attempt to bring patients’ and caregivers’ beliefs about each of these illness characteristics in line with the biomedical model (e.g., (Kosse et al., 2019; Petrie et al., 2012)).

However, asthma education interventions have shown limited success in reducing exacerbations (Morrison et al., 2014; Wolf et al., 2003). One reason is that a variety of factors at multiple levels of influence affect asthma management behavior and health outcomes, including factors outside the control of caregivers, children, and health care providers (Khan et al., 2021). A second possible reason is that, although researchers know much about what specific triggers and symptoms caregivers identify as important for raising their child’s risk of having asthma problems, they know little about how caregivers mentally represent asthma triggers and symptoms more generally. Mental representations can be thought of as multidimensional cognitive-affective structures that caregivers use to conceptualize and bring meaning to categories of objects (Fischhoff et al., 1978), such as triggers and symptoms. Furthermore, they underlie caregivers’ judgments about triggers and symptoms. Insights into these mental representations would allow us to identify the constellation of characteristics of triggers and symptoms that prompts caregivers to view their child as being at higher or lower risk of exacerbation. In other words, elucidating caregivers’ mental representations could help researchers understand what it is about triggers and symptoms that make them seem more or less risky.

Without information about caregivers’ mental representations of asthma triggers and symptoms, interventions that attempt to teach caregivers to avoid triggers and to treat early-stage symptoms such as fatigue—including interventions based on the Common Sense Model—may fail to convey the information in a way that makes sense to the caregiver, thereby reducing the effects of the intervention. In addition, caregivers likely can identify any number triggers and symptoms that are unrecognized by the biomedical model. This could lead to a situation whereby every time scientists develop an intervention to address one trigger or symptom, a new trigger or symptom emerges that requires a new intervention. In contrast, identifying a small number of categories that describe triggers and symptoms in general could make asthma education more manageable and provide insight for health care providers who may not understand why their patients are not following standard trigger and symptom management guidelines.

Identifying Mental Representations

Fischhoff and colleagues (1978) developed the psychometric paradigm of risk perceptions to explore the dimensions that structure laypeople’s subjective perceptions of technological, behavioral, and other hazards (e.g., firearms, tornadoes, nuclear power plants). In this paradigm, researchers employ data reduction methods such as multidimensional scaling and other multivariate analysis techniques to generate quantitative representations or “mental maps” of people’s perceptions of objects in a domain. Principal components analysis (PCA) is a common analytical tool in research with the psychometric paradigm for reducing data into a small number of meaningful dimensions. To identify people’s mental representations for a given topic domain (e.g., asthma symptoms), participants rate a variety of domain-related objects (e.g., wheezing) on several characteristics (e.g., controllable). Responses are analyzed in such a way as to produce linear combinations of the objects and characteristics. The resulting axes, or components, reflect commonalities in characteristics among the objects in the domain (Stevens, 2002).

Researchers have used PCA to extract the key dimensions structuring people’s mental representations of objects in several topic domains. Subsequent to Fischhoff et al.’s (1978) foundational work, researchers used PCA or its cousin exploratory factor analysis (EFA) to identify the cognitive-affective structures that underlie people’s mental representations of medication side effects (Waters et al., 2017), food products (Perkovic et al., 2022), prescription medications (Slovic et al., 2007), energy technologies (Bronfman et al., 2008), and automobile safety strategies (Slovic et al., 1987).

Researchers have also examined how mental representations are related to key outcomes, including perceptions of risk (Bronfman et al., 2008; Fischhoff et al., 1978; Slovic, 1987; Slovic et al., 1980; Slovic et al., 1987), perceived healthiness of food (Perkovic et al., 2022), and aversion to medication side effects (Waters et al., 2017). For example, in one study, PCA revealed four dimensions in people’s mental representations of medication side effects: affectively challenging, socially challenging, physically challenging, and familiar (Waters et al., 2017). Side effects that were viewed as more affectively and physically challenging were also perceived as more aversive (i.e., participants viewed the medication more negatively and as less desirable, they were less willing to take the medication, and they were willing to pay more money to avoid the side effect). These insights provided fertile ground for generating testable hypotheses that can be tested in real-world settings. For example, medications whose side effects pose affective challenge, such as depression, may have lower patient adherence than medications whose side effects are less affectively challenging.

Study Objective and Research Questions

Our objective was to gain insight into the dimensions underlying caregivers’ mental representations of their child’s asthma triggers and symptoms, with the ultimate goal of improving asthma education interventions. Consistent with other studies using the psychometric paradigm (e.g., (Fischhoff et al., 1978)), we used PCA as a primary analytic framework. We were guided by the following research questions:

1. What are the key dimensions that structure caregivers’ mental representations of asthma triggers and symptoms?

2. How are the dimensions that structure caregivers’ mental representations of asthma triggers and symptoms related to their interpersonal risk perceptions (i.e., their perceptions of their child’s risk of having an exacerbation)?

We pre-registered the research questions and approach. The pre-registration, study materials, data, and analysis code are located at the Open Science Framework: https://osf.io/awqf8/

Methods

Participants

The Institutional Review Board at University of Florida approved all procedures and measures (approval #201802313). We recruited participants from the Qualtrics market research survey panel of U.S. residents. Data were collected from September through December 2020. Eligibility criteria included: reporting having a child younger than age 18 who was diagnosed with asthma by a health care provider, and who still had asthma at the time of the survey. Caregivers whose children resided with them for fewer than 90 days a year were excluded. We stratified recruitment to attempt to recruit at least 2/3 of a sample with household incomes of less than USD $50,000, because low income is associated with increased asthma incidence and severity (Centers for Disease Control (CDC), 2022). Although we were unable to meet this goal, we were able to recruit a sample with approximately 52% with household incomes of less than $50,000.

Of 220,238 emails that Qualtrics sent to potentially eligible participants, 2,363 individuals clicked the survey link. Of those, 91 caregivers did not provide consent, 609 were excluded due to reporting an income of greater than $50,000, and 1,663 consented. We removed data from 145 caregivers who discontinued the survey after consenting, 283 who quit before completing key socio-demographic measures, 11 whose child was age 18 or older (or who did not report their child’s age), and 3 who reported being younger than age 18 themselves. Of the 1,221 remaining caregivers, we excluded 408 who failed at least two of five indicators of poor data quality and five who Qualtrics flagged as potential duplicates. The final analytic sample size was 808. Of those 808 caregivers, 190 were assigned to the triggers condition, 187 were assigned to the symptoms condition, 222 were assigned to the trigger management condition, and 209 were assigned to the symptom management condition. See Online Supplemental Materials Section A for a full discussion of the data quality decisions and Table 1 for the socio-demographic characteristics of caregivers and their children.

Table 1.

Caregiver and Child Socio-demographic Information

	N = 377	%
Caregiver Gender
Male	172	45.6
Female	204	54.1
Transgender	1	0.3
Caregiver Race a
White	306	78.7
Black or African American	50	12.8
Asian	10	2.6
Native American/Alaskan Native	6	1.5
Hawaiian/Pacific Islander	2	0.5
Multiracial/Other	12	3.1
Don’t know	1	0.2
Choose not to respond	2	0.5
Caregiver Ethnicity
Non-Hispanic	328	87.0
Hispanic	45	11.9
Don’t know	2	0.5
Choose not to respond	2	0.5
Caregiver Formal Education
Less than high school	15	4.0
High school graduate, GED, or high school equivalent	110	29.2
Vocational, trade, or technical school	25	6.6
Associate degree	54	14.3
Bachelor’s degree	77	20.4
Post-graduate degree	96	25.5
Choose not to respond	0	0.0
Household Income
$25,000 or less	141	37.4
$25,001 to $50,000	60	15.9
$50,001 to $75,000	62	16.4
$75,001 to $100,000	27	7.2
$100,000 or more	87	23.1
Choose not to respond	0	0.0
Financial Security b
Canť make ends meet	81	21.5
Manage to get by	171	45.3
Enough money to manage, plus extra	85	22.5
Money is not a problem	25	6.6
Choose not to respond	2	0.5
Missing	13	3.4
Home Ownership
Rent home	150	39.8
Own home	199	52.8
Some other arrangement	14	3.7
Don’t know	1	0.3
Missing	13	3.4
Child Gender
Male	210	55.7
Female	164	43.5
Transgender	1	0.3
Gender non-conforming	1	0.3
Choose not to respond	1	0.3
	M	SD
Subjective Social Status c [range: 1 (worst off) to 10 (best off)	5.3	2.6
Caregiver Age [range: 18 to 60]	37.5	8.3
Child Age [range: 0 to 17]	9.9	4.2
Control of asthma symptoms d [range: 4 (low) to 20 (high)]	12.9	4.7
Subjective asthma control e [range: 1 (low) to 5 (high)]	3.9	0.9

^aDoes not sum to 377because categories are not mutually exclusive.

^bWe measured financial security using a single item: Which of these statements best describes your present financial status? 1 (can’t make ends meet), 2 (manage to get by), 3 (enough to manage plus some extra, and 4 (money is not a problem).

^cWe measured subjective Social Status using the MacArthur ladder (Adler et al., 2000).

^dWe measured asthma control using the sum of items 1–4 of the parent proxy asthma control test (PP-ACT) (duRivage et al., 2017).

^eWe measured subjective asthma control using item 5 of the PP-ACT (duRivage et al., 2017).

Design

We used the psychometric paradigm (Fischhoff et al., 1978) to examine caregivers’ mental representations of four broad domains of asthma concepts: triggers, symptoms, trigger management strategies, and symptom management strategies. Accordingly, caregivers rated several objects (e.g., dust) in a domain (e.g., triggers) across multiple characteristics (e.g., detectable). For this study, we obtained caregivers’ ratings of 20 objects along 15 characteristics for each of the four categories (specifically, triggers, symptoms, trigger management behaviors, and symptom management behaviors). This resulted in 1200 total ratings for the entire study (i.e., 4 categories x 20 objects x 15 characteristics). In this paper, we describe only items related to the categories asthma triggers and symptoms. Results describing the mental representations of asthma trigger and symptom management behaviors will be reported separately.

To limit burden, we asked caregivers to rate only 10 of the 80 total objects under study. Specifically, we randomly generated two sets of ten objects for each of the four categories. Caregivers were randomly assigned to one of the eight sets, and they rated only the 10 objects in their assigned set across 15 characteristics. This resulted in 150 ratings per caregiver (1 set of 10 objects x 15 characteristics). This approach ensured that each object had a similar number of responses.

Materials

We engaged in a multistep process, grounded in the literature, to determine which characteristics and objects should be included in the PCA. We began by developing an initial pool of objects by reviewing the biomedical literature about asthma (Busse et al., 2012; Krishnan et al., 2012; National Heart Lung and Blood Institute, 2007), seeking advice from asthma experts, and examining findings from our qualitative interviews of caregivers of children with asthma and their health care providers (Hunleth et al., 2020; Spray et al., 2022). We developed an initial pool of characteristics by reviewing the literature on risk perception (including the role of affect and emotion in risk perception) (Finucane et al., 2000; Lerner & Keltner, 2000; Loewenstein et al., 2001; Slovic, 1987; Slovic et al., 1980; Waters et al., 2017), and the literature on illness and symptom representations (Bishop, 1987; Leventhal et al., 2003). We also created additional characteristics based on our qualitative research (Hunleth et al., 2020; Spray et al., 2022). We then conducted a pilot study to identify which objects and characteristics showed little variability and therefore could be excluded. We also sought to maximize conceptual breadth and minimize conceptual redundancy. See the Online Supplemental Materials Section B, Tables 1–3 for the exact wording for all objects and characteristics.

Table 3.

Results of the mixed-effects regression model analysis predicting caregivers’ interpersonal perceived risk based on the dimensions underlying their mental representations of asthma triggers and symptoms. Credible effects are in boldface. 95% highest density intervals (HDIs) are in brackets.

Type of object	Term	Estimate	CI95%
Triggers
	Intercept	0.06	[−0.08, 0.21]
	Severe and relevant	0.79	[0.74, 0.85]
	Negative affect—yet manageable	0.13	[0.07, 0.19]
	Chronic—yet unpredictable	0.01	[−0.04, 0.07]
Symptoms
	Intercept	0.22	[−0.05, 0.50]
	Severe and unpredictable	0.85	[0.78, 0.92]
	Relevant and common	0.09	[0.03, 0.14]
	Manageable despite unknown cause	−0.03	[−0.10, 0.03]
	Self-blame	0.04	[−0.01, 0.09]

Note. Estimates are non-standardized coefficients. Because in the error variance in mixed-effects modeling comes from multiple sources, it is not appropriate to standardize coefficients by their standard errors.

Interpersonal risk perceptions.

Because many caregivers were unfamiliar with the term “exacerbation,” we first described its characteristics and labeled it an “asthma episode.” For caregivers randomly assigned to the trigger domain, we first asked whether a particular trigger object had ever caused their child to have an asthma episode (yes, no, don’t know). In both the trigger and symptom categories, caregivers reported the riskiness of the trigger/symptom for their child’s asthma (1 = not at all risky; 5 = extremely risky). Caregivers could also choose “don’t know” or “choose not to respond.” These responses comprised only 0.30% of all responses for triggers, and only 0.33% of all responses for symptoms. See Online Supplemental Materials Section B, Table 3 for the full text and response options.

Objects.

The final list of 20 trigger objects included a variety of allergens, irritants, physiological and psychological states, and environmental factors. The final list of 20 symptom objects included several symptoms of illness, allergies, and symptoms that can arise because of asthma. The objects varied in the extent to which they were supported by the biomedical model of asthma, because we wanted to ensure a comprehensive understanding of caregivers’ mental representations. See Online Supplemental Materials Section B, Tables 1–2.

Table 2.

Results of the principal component analyses. Loadings of characteristics on each component for triggers and symptoms.

Characteristic		Triggers			Symptoms
		Severe and relevant	Negative affect—yet manageable	Chronic—yet unpredictable	Severe and unpredictable	Relevant and common	Manageable despite unknown cause	Self-blame
Worried		0.97	0.03	−0.23	1.02	−0.11	0.07	−0.03
Angry		0.20	0.85	−0.06	n/a	n/a	n/a	n/a
Annoyed		0.38	0.74	−0.05	−0.21	0.23	0.41	0.68
Self-blame		0.04	0.93	0.00	0.11	−0.09	−0.11	0.97
Unpredictable		0.42	0.16	0.60	0.56	0.20	0.11	0.35
Immediate		0.88	0.22	0.09	n/a	n/a	n/a	n/a
Limiting		0.94	−0.08	0.10	0.85	0.05	0.16	0.10
Exhausting		0.94	0.05	0.06	n/a	n/a	n/a	n/a
Serious		0.96	0.10	−0.08	0.98	0.00	0.00	0.02
Chronic/persistent		0.23	−0.35	0.76	−0.16	1.03	0.03	0.02
Relevant		0.95	−0.24	0.10	0.34	0.78	0.03	0.01
Common		0.82	0.15	0.17	0.41	0.72	0.04	−0.02
Detectable		−0.15	0.90	0.47	0.78	0.14	−0.38	0.05
Uncomfortable		n/a	n/a	n/a	0.81	0.02	0.26	0.10
Indicates asthma episode		n/a	n/a	n/a	0.75	0.36	−0.01	0.04
Uncertain		n/a	n/a	n/a	0.16	−0.22	0.71	0.29
Avoidable		−0.01	0.86	−0.30	n/a	n/a	n/a	n/a
Preventable		n/a	n/a	n/a	0.24	0.31	0.41	0.35
Controllable		0.03	0.82	−0.24	0.14	0.24	0.82	−0.06
Variance explained	Individual components	51%	37%	12%	44%	12%	16%	16%
	Total		87%		91%

Note. n/a indicates that the characteristic was not assessed for the domain depicted by the column (e.g., participants did not rate the extent to which trigger objects were uncomfortable).

Characteristics.

The final list of 15 characteristics varied based on the domain of objects. Eleven characteristics overlapped for the trigger and symptom categories: worried, self-blame, limiting, unpredictable, detectable, controllable, serious, annoyed/frustrated, common, chronic/persistent, relevant. Four characteristics were unique to the triggers (avoidable, angry, immediate, and exhausting) and four were unique to symptoms (preventable, uncomfortable, uncertain, and indicates an asthma episode). Caregivers rated each object using a 5-point scale (1 = do not agree at all; 5 = agree completely). See Online Supplemental Materials Section B, Table 3 for a list of all characteristics.

Socio-demographic characteristics.

We assessed several socio-demographic characteristics that are related to asthma outcomes in children in the U.S. (see Table 1).

Procedure

After providing consent, we asked caregivers to report how many children with asthma they had and the age and gender of their youngest child with asthma. Then, we randomly assigned caregivers to view one of the eight sets of 10 objects. For each object, caregivers indicated their perceived likelihood of their child having an asthma exacerbation (i.e., interpersonal perceived risk). Next, they rated the object according to each of the 15 characteristics. This process of providing their interpersonal perceived risk and rating objects across characteristics repeated for five objects. Then, to provide caregivers a break from the rating task, caregivers provided their own age, gender, race, ethnicity, level of formal education, and asthma status before rating the characteristics for the remaining five objects in the set. Finally, caregivers provided additional socio-demographic information (i.e., financial security and subjective social status) and indicated their personal tobacco use status, their child’s exposure to secondhand smoke, and an item evaluating self-reported attention to the survey.

Within a set, we randomized the order of the 10 objects we presented. Caregivers could choose to skip any item. The median completion time was approximately 25 minutes.

Results

We began by calculating the average (across caregivers) interpersonal perceived risk of asthma exacerbation for each trigger and symptom. As shown in Figure 1, although interpersonal risk perceptions were moderate-to-high for most triggers and symptoms, there was also variability in the extent to which caregivers perceived a certain trigger or symptom as risky. For example, the triggers illness of the lungs, tobacco smoke, mold, and pollen elicited the highest interpersonal risk perceptions, whereas pests elicited the lowest interpersonal risk perceptions. For symptoms, discolored face or hands, chest or throat retractions, wheezing, and shortness of breath elicited the highest interpersonal risk perceptions, and posture changes, throat feeling itchy, sneezing, becoming quiet, snoring, and needing comfort elicited the lowest interpersonal risk perceptions.

Figure 1. Boxplots showing the distribution of interpersonal risk perception ratings for asthma triggers (left-hand graph) and for asthma symptoms (right-hand graph).

Note. Triggers and symptoms are ordered by the average, across caregivers’ risk rating from high (at the top of the graph) to low (at the bottom of the graph). The grey bars indicate the interquartile range (IQR). The broken horizontal lines indicate the range of the responses (+/− 1.5*IQR), and the dots indicate outliers.

What are the key dimensions that structure caregivers’ mental representations?

Following previous applications of PCA in studies using the psychometric paradigm (e.g., (Fischhoff et al., 1978; Perkovic et al., 2022; Waters et al., 2017), we first calculated the average (across caregivers) rating for each object on each characteristic. Then, separately for triggers and symptoms, we used these average ratings to conduct a PCA using the principal function from the psych package (Revelle, 2019) in R (R Core Team, 2020). We decided how many components to retain for each domain of objects by using the Kaiser criterion (i.e., the requirement of eigenvalues being larger than 1) and the “elbow” criterion (i.e., the point in the scree plot where the eigenvalues of additional components level off). We used oblimin rotation to facilitate interpretation of the solutions. Table 2 depicts: (a) how each individual characteristic loaded on the principal components for the triggers and symptoms, (b) the percentage of variance explained by each of the retained components, and (c) the total variance explained by each model. Figure 2 shows show how individual objects scored on the extracted components.

Figure 2. Two-dimensional factor space depicting caregivers’ mental representations of (A) triggers and (B) symptoms on the components extracted by the principal component analyses.

Note. The label on the bottom of each factor space represents the horizontal axis (e.g., “severe and relevant” located on the left-hand side of panel A). The label to the left of each factor space represents the vertical axis (e.g., “negative affect—yet manageable” located on the left-hand side of panel A).

For triggers, we retained three components (see Table 2). For the first component, the highest loadings were for the characteristics of “limiting,” ”serious,” and ”exhausting,” ”relevant,” and ”common.” We interpreted this component as severe and relevant. Triggers scoring high on the component were illness of the lungs and physical activity; triggers scoring low were pests and pets (see Figure 2, Panel A). The highest loadings on the second component were for ”angry” and ”self-blame,” but also ”avoidable” and ”controllable;” we thus refer to this component as negative affect—yet manageable. Examples of triggers with high scores on the component were tobacco and mold; dry air and pollen had low scores. On the third component, the highest loadings were for ”unpredictable” and ”chronic/persistent;” we thus refer to it as chronic—yet unpredictable. Triggers scoring high on this component were change of season and strong smells; triggers with low scores were pests and stress.

For symptoms, we retained four components (see Table 2). For the first component, the highest loadings were for ”worried,” ”serious,” ”limiting,” and ”unpredictable;” we interpreted this component as severe and unpredictable. Symptoms scoring high on the component were discolored facial features or fingernails and tightness in the chest; symptoms with low scores were need for comfort and snoring (see Figure 2, Panel B). Characteristics with high loadings on the second component were ”chronic/persistent,” “common,” and ”relevant;” we thus refer to it as relevant and common. Characteristics with high loading on the third component were “uncertain,” ”controllable,” and ”preventable,” and we thus refer to it as manageable despite unknown cause. Itchy throat and dry cough are examples of symptoms with high scores on this component; snoring and need for comfort are examples for symptoms with low values. Finally, characteristics with high loadings on the fourth component were ”self-blame” and ”annoyed,” and we thus referred to it as self-blame. Waking up at night and vomiting had high scores on this component; snoring and fever had low values.

How are scores on the components associated with interpersonal risk perceptions?

We next examined the extent to which the extracted components underlying caregivers’ mental representations of triggers and symptoms were associated with their interpersonal risk perceptions of the triggers and symptoms. We used Bayesian mixed-effects modeling using the brms package in R (Bürkner, 2017), with the individual caregivers’ responses as the unit of analysis (instead of aggregating across caregivers, as in the PCA analyses). Although each caregiver responded to only a subset of characteristics and objects, the mixed-effects approach is robust to missing data. For making inferences about the link between the components and the risk perceptions, we relied on the population (i.e., fixed) effects. Specifically, we used the different components as fixed effects and included random intercepts for caregivers and the different objects (i.e., triggers and symptoms). To compute the score for an object on a given component for a given caregiver, we averaged the caregiver’s responses for those characteristics that loaded highest on that component (as shown in bold in Table 2) and for which the caregiver actually provided a response.

Table 3 shows the estimated regression coefficients for the fixed effects of the extracted components when predicting individual caregivers’ interpersonal risk perception from trigger and symptom ratings. Credible effects (i.e., coefficients whose 95% confidence interval did not include zero) are bolded. Triggers that obtained high ratings on the severe and relevant component, as well as on the negative affect—yet manageable component, were associated with higher interpersonal perceived risk ratings. Symptoms that obtained high ratings on the severe and unpredictable component, as well as on the relevant and common component, were associated with higher risk ratings.

Discussion

This study demonstrates the richness and complexity in the way caregivers conceptualize their child’s asthma triggers and asthma symptoms, and how these conceptualizations guide the perceived risk of experiencing an asthma exacerbation. Although other studies have described caregivers’ beliefs about asthma triggers, symptoms, and risk, in many (but not all) cases, their goals focus on evaluating the extent to which caregiver-identified triggers and symptoms are consistent with the biomedical model of asthma (Sonney et al., 2016) (for counterexample, see (Peterson et al., 2002)). Instead, this study took a different methodological and conceptual approach by drawing on the psychometric paradigm to explore the multidimensional nature of caregivers’ mental representations of asthma triggers and symptoms, and how those dimensions are related to their interpersonal perceptions of risk. This approach extends prior asthma research by providing insight about what characteristics of asthma triggers and symptoms make them seem more and less risky, rather than simply identifying beliefs about individual triggers and symptoms. As described below, such insights could make asthma education interventions more manageable and effective and provide insight for health care providers, who may not understand why their patients are not following standard trigger and symptom management guidelines.

The findings show several similarities between caregivers’ mental representations of triggers and their mental representations of symptoms. Specifically, mental representations of both triggers and symptoms included dimensions that reflected the extent to which caregivers perceived them as severe, manageable, unpredictable, or relevant to their child, or they elicited negative affect in the caregiver (including self-blame). These results align with several elements of the Common Sense Model (CSM); for example, severe, manageable, unpredictable, and negative affect are similar to the CSM’s concepts of consequences, control, identity, and emotional representation, respectively (Broadbent et al., 2006; Leventhal et al., 2003). This converging evidence from different conceptual and methodological approaches lends credence to the importance of these concepts.

Personal Responsibility, Self-Blame, and Responsibilization

More importantly than where the components of this study overlap with elements of the CSM, however, is what they tell us about how seemingly disparate characteristics overlap. Take, for example, the overlap between the components negative affect—yet manageable (for triggers) and self-blame (for symptoms). Understanding this overlap requires recognizing that the feelings people experience occur within a modern US healthcare system that privileges individual responsibility (Mulligan & Castañeda, 2018). In brief, the US healthcare system places pressure on individuals to be self-reliant, self-regulating, and forward thinking (for more on neoliberalism and individual responsibility, see (Clarke, 2005; Lemke, 2001; Merry, 2009; Rose, 1999) as cited in (Hunleth et al., 2016). This expectation of responsibility, what scholars have referred to as responsibilization, is evident in how the labor for health rests on the individual and their caregivers (Horton et al., 2014). It is further evident in public health discourse, including in the context of “making healthy lifestyle choices” to prevent chronic disease, and “taking care of yourself” by engaging in preventive screening activities and adhering to treatment guidelines. Critically, the “judgments about what people ought to do, which are implicit in these discourses, can become internalized in patients and caregivers. They affect people’s practices and behaviors and their perceptions of themselves” (Hunleth et al., 2016). Consequently, when people become ill, such judgments can lead to feelings of self-blame, shame, and guilt (Griffiths et al., 2006; Hunleth et al., 2016; McMullin et al., 2009). This suggests that, although feelings of self-efficacy (e.g., feeling agentic or capable of taking action) can be helpful in promoting certain behaviors (Bandura, 2004), feelings of agency may also become harmful when agency is constrained or does not lead to desired outcomes. In the specific context of pediatric asthma caregiving, the responsibility has been placed on the caregiver to ensure that their child remains healthy (Klok et al., 2011). Thus, any symptoms could be interpreted by the caregiver as a failure to protect their child, and result in the aforementioned feelings of self-blame, shame, and guilt (Fawcett et al., 2019).

The negative affect—yet manageable component includes characteristics that were framed in ways that might have tapped into caregivers’ feelings of and expectations about individual responsibility demanded of them within the US healthcare context. Specifically, we asked caregivers not only how detectable, controllable, and avoidable a trigger was, but also how much self-blame, anger, and frustration they would feel if their child experienced a certain trigger—In other words, how would they feel if they “failed” to manage their child’s trigger exposure. When considered alongside the severe and relevant trigger component, it is reasonable to imagine a caregiver feeling like they are responsible for exposing their child to a dangerous substance (although we did not ask caregivers to indicate whether their anger and frustration was directed toward themselves or toward other people, the fact that those characteristics loaded on the component alongside self-blame, rather than a different component, suggests it is more likely that such feelings were directed inward than outwards). We see comparable findings for the components underlying mental representations of symptoms—as severe and manageable, but also eliciting feelings of self-blame when they occur. Although we did not assess general levels of asthma anxiety in the caregivers, it is important to note that caregivers’ feelings of anxiety about the unpredictable nature of asthma can produce heightened vigilance (Fawcett et al., 2019). In addition, asthma management responsibilities are labor-intensive, stressful, and financially burdensome. It is therefore straightforward to imagine a situation in which a caregiver who is struggling with anxiety and with balancing competing responsibilities might feel guilt, shame, and self-blame when they perceive that they have “failed” to prevent an exacerbation.

This discussion about responsibility and self-blame is also relevant for the findings that are not associated with self-blame. Specifically, although the trigger component negative affect—yet manageable includes the characteristic self-blame and is associated with higher interpersonal perceived risk of exacerbation, the symptom components self-blame and manageable despite unknown cause are not related to interpersonal risk perceptions. Instead, the components that are most strongly associated with interpersonal risk perceptions are those related to severity – severe and relevant for triggers, and severe and unpredictable for symptoms. The association between severity and interpersonal risk perceptions are consistent with an extensive body of literature indicating that severity is a key element of personal risk perceptions (Janz & Becker, 1984; Slovic et al., 1980). Also notable is that worry was the attribute that most strongly loaded on the severity-related components. In addition, the severity-related components were the components that related most closely to interpersonal risk perceptions. These patterns of data are consistent with extensive research describing the close relationship between perceptions of risk and negatively-valenced avoidance-related emotions like worry, fear, concern, and dread (Ferrer et al., 2016; Fischhoff et al., 1978; Lerner & Keltner, 2000; Loewenstein et al., 2001; Slovic, 1987).

What are the practical implications of mental representations that reflect responsibilization if such reflections (e.g., self-blame) are only inconsistently related to interpersonal risk perceptions? Responsibilization places the onus of a societal problem on the individual. However, in the U.S., asthma incidence and outcomes—particularly disparities in asthma incidence and outcomes—originate in large part from economic constraint and health insurer policies that limit the affordability of preventer medications, racial discrimination and residential segregation that limits trust in and accessibility to health care providers, lax zoning laws that do not sanction landlords for not remediating mold and pest infestations, and environmental injustices that expose marginalized communities to air pollution (Harris, 2019; Hunleth et al., 2020; Ruiz et al., 2022). Mental representations that reflect responsibilization by prioritizing individual action risk shifting focus away from the root causes of asthma triggers and symptoms and onto caregivers who, in many cases, already experience multiple vulnerabilities. This, in turn, could lead to exacerbation of asthma disparities and increased burden on already-overburdened caregivers.

Strengths, Limitations, and Future Directions

This research had several strengths. First, we continue to extend the use of PCA into new areas, demonstrating the utility of the method and analytic framework for understanding mental representations of different topic domains. Second, whereas much research that has used PCA to understand mental representations focuses on a broad array of hazards in different domains, we focus on understanding mental representations of two key elements of caregivers’ understanding of their child’s asthma. Our focus allows us to gain deep understanding of a single topic that is critical to children’s health. Third, this research was conducted in a large sample of caregivers from around the U.S. This sampling strategy provides results that are more geographically generalizable than research that is conducted in only a single clinic or in a handful of clinics.

This study’s results should be considered in the context of several limitations. First, like most research examining mental representations of health hazards, this study relied on observational data and a cross-sectional research design. Thus, rather than caregivers’ mental representations structuring their interpersonal risk perceptions, it is possible that caregivers use their interpersonal risk perceptions to construct their mental representations. Second, due to the structure of the survey design, we were unable to determine whether mental representations affected asthma management behavior through interpersonal risk perceptions. Third, because we were concerned about overburdening participants with a time-consuming and repetitive task, we asked caregivers to rate either triggers or symptoms. This decision precludes our ability to compare a single caregiver’s mental representations of triggers to their mental representations of symptoms. Fourth, due to resource constraints, we were only able to ask these research questions in the context of asthma. Children experience many chronic health conditions that require management and support from caregivers, including Type I diabetes, sickle cell disease, food allergies, epilepsy, and juvenile rheumatoid arthritis; we are unable to draw conclusions about the mental representations of for these diseases, although we can draw hypotheses (e.g., severe is likely to be one of the dimensions). Fifth, although our sample was generally similar to the broader US population in terms of the proportion of individuals from different racial and ethnic backgrounds (except for underrepresentation among Asian and Hispanic participants) (U. S. Bureau of the Census, 2023), including more Black, American Indian, and Hispanic caregivers would have lent additional assurance that the results of this study were applicable to caregivers who may experience the most challenges managing their child’s asthma. It is also notable that these data were collected in the first months of the COVID-19 pandemic. Many studies found improved asthma control and fewer hospitalizations for pediatric asthma, and caregivers who perceived that their children’s asthma control was either unchanged or improved (81%) compared to before the pandemic (Pogge et al., 2022). If that was the case for caregivers in this study, they may have rated some of the objects (e.g., inhaler use, doctor visit) as not relevant for their child (whereas they may have marked them as relevant prior to the pandemic). Unfortunately, we do not have data available to explore how the pandemic might have affected caregivers’ beliefs about asthma triggers, symptoms, or exacerbation risk.

Research that addresses the limitations just described will likely yield important theoretical and practical advances. For example, research designs that explore the causal and directional relationships among mental representations, interpersonal risk perceptions, and asthma management behaviors could either enhance confidence in theories that assert such relationships exist or demonstrate their need for revision. For example, although the psychometric paradigm of risk perception (Slovic et al., 1980) may suggest hypotheses about the relationship between certain dimensions of mental representations and perceived risk, we know of no research that has examined whether risk perceptions mediate the relationship between mental representations and behavior. Such research in the context of asthma triggers and symptoms would provide valuable support for—or against—the practical validity of the psychometric paradigm. In addition, demonstrating whether these relationships are consistent across different child health conditions (e.g., sickle cell disease, Type 1 diabetes, etc.) could provide valuable information about the specificity or generality of theories that include mental or illness representations as elements. Finally, this same methodology could be applied to an examination of adolescents’ mental representations of their own triggers and symptoms, and how those representations affect their own risk perceptions. It is possible that the rating activity itself could help adolescents gain insight into their condition, which could help them transition from pediatric to adult care settings.

Implications for Research and Practice

Our exploration suggests several possibilities for improving asthma education, health care provider communication, and, ideally, asthma outcomes. First, we demonstrated that caregivers view triggers and symptoms as highly-laden with negative affect, directed both outward toward the child (e.g., worry) and inward toward the caregiver (self-blame). Asthma education interventions and health care provider-initiated conversations that prepare caregivers for experiencing these feelings and provide coping strategies may be particularly credible and useful. Interventions might also help caregivers identify which triggers and symptoms they are most (and least) aware of, concerned about, and responsive to. Second, the names of the dimensions underlying mental representations of triggers and symptoms suggest possible content for interventions. Severe, unpredictable, manageable, and relevant are all descriptors of components of both triggers and symptoms. Thus, for example, an intervention could help caregivers and children learn to predict when triggers or symptoms will affect their child, or, if that is not possible, to help caregivers and children cope with the aftermath. Third, these results could be used to help calibrate caregivers’ risk perceptions to their child’s health condition. Educational interventions or provider conversations that emphasize—or deemphasize, as appropriate—the severity, relevance, manageability, or predictability of triggers and symptoms would likely also increase or decrease caregivers’ risk perceptions. However, it is important to avoid inadvertently blaming caregivers who are unable to overcome powerful structural forces that limit the extent to which personal action can improve health. It is also important to note that any efforts to implement these interventions should be preceded by research intended to test their efficacy, appropriateness, and safety.

Complex diseases such as pediatric asthma present significant challenges to caregivers tasked with helping the child managing the disease. Our study provides insights into the key components underlying caregiver risk perceptions and suggests paths for education and intervention.

Data availability:

The de-identified data, the analytic code used to conduct the analyses presented in this manuscript, and all materials used to conduct the study are freely available in the Open Science Framework public archive: https://osf.io/awqf8/?view_only=b965206c97684ab094e96e960826bf63