Psychological Traits and the Persuasiveness of Lung Cancer Screening Health Messages

Abstract

Background:

Lung cancer screening remains underutilized despite its proven mortality benefit. Health systems have attempted to increase screening awareness through advertising. Psychological theories suggest that construal level (a personal orientation towards the big picture or the details) and regulatory focus (goals emphasizing acquisition of a good, or avoidance of a bad outcome) play a key role in health advertising effectiveness. These theories have not been examined in a screen-eligible population.

Methods:

Using Amazon’s crowdsourcing platform, Mechanical Turk, we identified screen-eligible individuals based on United States Preventive Services Task Force criteria. We randomly assigned participants to see one of four screening advertisement images in a 2 (construal level: high vs. low) X 2 (regulatory focus: promotion vs. prevention) between-subjects experimental design. We assessed willingness to undergo screening following the advertisement.

Results:

A total of 191 individuals responded to our study invitation (mean age 61 years). We found that the high construal/promotion focus image led to a greater willingness to screen compared to images representing other psychological states (p-value=0.04). Regarding the personality traits of our respondents, high construal/promotion focus was the most prevalent (40%) trait combination, whereas low construal/prevention focus was the least prevalent (17%).

Conclusions:

The psychological focus of health-related messages affect an individual’s willingness to undergo lung cancer screening. Individuals eligible for lung cancer screening are more persuaded by “big picture” messages describing the benefits of screening. Health systems may use this knowledge to design more effective patient-facing communications that lead to higher rates of screening.

Graphical Abstract

Lung cancer screening saves lives^1,2 but is greatly underutilized, occurring in less than 12-18% of eligible individuals^3,4. One barrier to screening is a lack of awareness of screening among the eligible population^5,6. Health systems and patient advocacy groups have attempted to increase awareness through media campaigns and advertisements for lung cancer screening^7–9. It is not known whether some messages are more effective than others.

Psychologists classify messages by how concretely or abstractly they present information^10,11. If a message is more abstract or focused on the “big picture” it is considered high construal. Conversely, more concrete or detail-oriented messages are characterized as low construal. Another aspect of messages is the type of strategy highlighted in the message; one framework psychologists utilize for this purpose is regulatory focus¹¹. There are two types of regulatory focus: promotion focus emphasizes the pursuit of gains and aspirations toward ideals whereas a prevention focus emphasizes the avoidance of losses and the fulfillment of obligations.

Two psychological theories suggest that construal level and regulatory focus can affect the persuasiveness of a message. The first theory predicts that messages with pairs of characteristics—specifically high construal and promotion-focused or low construal and prevention-focused—are most persuasive in a general population¹⁰. A second theory relates to construal level and regulatory focus as traits that individuals possess, which can be measured with standardized instruments^13,14. This theory suggests that an alignment between a message’s characteristics and the traits of the individual viewing the message increases the message’s persuasiveness^10,12,15. These theories have been supported by studies in various consumer research settings but not been examined in a healthcare context involving high-risk subpopulations.

Our first goal was to determine if advertisement images (characterized by different combinations of construal level and regulatory focus) have an effect on the self-reported likelihood of undergoing lung cancer screening. Our second goal is to determine if alignment between an image’s characteristics and the traits of the individual is associated with an increased likelihood of screening. We hypothesize that different types of images have an effect on the self-reported likelihood of undergoing screening and that alignment is associated with increased likelihood of screening.

Material and Methods

Participant Population

We previously conducted a study on lung cancer screening knowledge, attitudes, and beliefs among screen-eligible individuals using Amazon’s crowdsourcing platform, Mechanical Turk (MTurk)⁶. MTurk is a crowdsourcing recruitment tool for survey research^16,17. Screen eligibility was defined by United States Preventive Services Task (USPSTF) criteria (age 55-80 years, former or current 30 pack-year smokers, former smokers who quit within the last 15 years)¹⁸. In addition to smoking history data, we collected demographic, clinical, and other lung cancer risk-factor information. All two hundred forty individuals from that study were invited to participate in this study. We offered participants $4 as compensation. All study materials were approved by the University of Washington Institutional Review Board (STUDY00003234) and informed consent was waived.

Study Design

Using the proprietary built-in “Question Randomization” feature of an online survey platform Qualtrics, we randomized participants to view a lung cancer screening advertisement image consistent with one of four psychological traits arising from a combination of two types of construal levels (high, low) and two types of regulatory focus (promotion, prevention). We then asked them about their likelihood to undergo lung cancer screening. Likelihood was measured on a five-point Likert scale with 1 representing “very likely to screen” and 5 representing “very unlikely to screen.” Because we could not engage individuals on MTurk to participate in shared decision making or screening, we used likelihood to screen as a surrogate for engaging in shared decision making and screening.

Image Selection

Our approach to image selection was pragmatic—we used Google’s search engine to conduct an image search for “lung cancer screening advertisement” and “lung cancer screening ad.” We downloaded the first 100 image results. After excluding images that did not have free-to-use licenses we were left with 18 images. Each image was viewed by five individuals in our research group and independently characterized by its construal level and regulatory focus. Our research group was composed of clinicians, health services researchers, and communication science experts from the University of Washington School of Business. Of the 18 images reviewed by the group, there was agreement on both the construal level and regulatory focus of eight. Of these, we selected four images to use in our study.

Personality Trait Determination

To achieve our secondary aims, we measured participant construal level and regulatory focus scores using standardized instruments—the Behavior Identification Form (BIF)¹⁴ and the Regulatory Focus Questionnaire (RFQ)¹³, respectively. The BIF determined if participants were of high or low construal level—referring to a preference for either “big picture” concepts or “detail-oriented” information, respectively. BIF scores range from 0 to 10; scores less than 5 represent a low construal and scores 5 or greater represent a high construal, based on the instrument developers’ recommended score interpretation¹². The RFQ determined if participants were promotion- or prevention-focused—signifying whether participants were primarily motivated by the prospect of gaining “good” versus preventing “bad” outcomes, respectively. RFQ scores range from −4 to +4; scores less than 0 represent a prevention focus and scores 0 or greater represent a promotion focus. This method of interpreting RFQ scores is recommended by the instrument developer and has been used in several studies^19–21. The four possible combinations of personality traits (high construal/promotion focus; high construal/prevention focus; low construal/promotion focus; and low construal/prevention focus) represent the same combination of characteristics of the four images we randomized participants to view (Figure 1).

Figure 1. Lung Cancer Screening Image Advertisements.

The high construal promotion focus image describes the big picture around lung cancer screening and focuses on potential benefits, such as increased survival rates with early-detection (image courtesy of the American Lung Association). The high construal prevention focus image provides a brief warning regarding lung cancer risk even after smoking cessation—at the top of the image—while screening is only mentioned near the bottom of the image. The low construal promotion focus image provides detailed information on lung cancer screening with an emphasis on its benefits. The low construal prevention focus image has large, lung-shaped ash trays featured prominently and focuses on the concern viewers might have about lung cancer while the potential benefits of screening are discussed in smaller font.

Study Deployment

We invited individuals to participate in our study over 14 weeks, between September and December 2018. After our initial invitation, we sent email reminders to individuals that had not yet completed our study every two weeks. We repeated this process until we went two weeks without obtaining additional participants. The study was completed using Qualtrics’ online survey platform. In order to control for data quality, we prevented respondents from participating more than once by using internet browser-specific “cookies” that blocked multiple attempts from the same internet device or user account.

Statistical Analysis

We reported means with standard deviations (SD) or medians with interquartile ranges (IQR) for normally distributed and non-normally distributed continuous variables, respectively. Frequencies with binomial exact confidence intervals (CI) were calculated for categorical variables. Mean likelihood scores were compared using analysis of variance (ANOVA). We used Kruskal-Wallis tests to evaluate differences in medians. All statistical analyses were conducted using R (version 3.4.4.).

Results

Of the 240 individuals invited to participate, 191 (79%) responded and completed our study. Table 1 shows the demographic, clinical, and risk-factor information of our participants. The median age was 60 years, the majority of participants were white and current smokers, had health insurance, at least a high school education, and income greater than $25,000 per year. Participants had a median 44 pack-year smoking history. Among former smokers, the median time since quitting was 6 years. There were no statistically significant differences across groups in terms of demographic, clinical, and risk-factor variables (Table 1). We compared the demographic, clinical, and risk-factor variables of the 49 individuals who did not respond to our study invitation, to our 191 study participants. Non-responders had significantly lower rates of reported COPD (6% vs 22%, p=0.02), but were otherwise similar to participants in our study (Supplemental Table 1).

Table 1.

MTurk Participant Characteristics by Image Type

Image type	All Participants (n=191) n (%, IQR)	High construal, Promotion focus (n=43) n (%, IQR)	High construal, Prevention focus (n=43) n (%, IQR)	Low construal, Promotion focus (n=49) n (%, IQR)	Low construal, Prevention focus (n=56) n (%, IQR)	p-Value
Age
Median	60 (7)	60 (5)	59 (5)	60 (7)	61 (8)	0.51
Women	132 (69%)	29 (67%)	30 (70%)	34 (69%)	39 (70%)	0.99
Race
White	174 (91%)	39 (91%)	39 (91%)	46 (94%)	50 (89%)	0.60
American Indian/Alaska Native	2 (1.0%)	1 (2.3%)	0 (0.0%)	0 (0.0%)	1 (1.8%)
Asian	2 (1.0%)	0 (0.0%)	0 (0.0%)	1 (2.0%)	1 (1.8%)
Native Hawaiian/Pacific Islander	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Black	8 (4.2%)	3 (7.0%)	2 (4.7%)	2 (4.1%)	1 (1.8%)
Other	6 (3.1%)	0 (0.0%)	2 (4.7%)	0 (0.0%)	3 (5.4%)
Hispanic	3 (1.6%)	0 (0.0%)	0 (0.0%)	1 (2.0%)	2 (3.6%)	0.62
Highest education
Less than high school	4 (2.1%)	2 (4.7%)	1 (2.3%)	1 (2.0%)	0 (0.0%)	0.47
High school graduate/GED	26 (14%)	6 (14%)	7 (16%)	6 (12%)	7 (13%)
Post-high school training, excluding college	46 (24%)	11 (26%)	13 (30%)	7 (14%)	15 (27%)
Associate’s degree or some college	65 (34%)	19 (44%)	12 (28%)	18 (37%)	16 (29%)
Bachelor’s degree	36 (18%)	4 (9.3%)	9 (21%)	11 (22%)	12 (21%)
Graduate school	14 (7.3%)	1 (2.3%)	1 (2.3%)	6 (12%)	6 (11%)
BMI
Median	27 (7)	28 (5)	27 (5)	26 (7)	25 (7)	0.07
Current Smoker	114 (60%)	29 (67%)	26 (61%)	33 (67%)	26 (46%)	0.09
Pack-years
Median	44 (13)	45 (18)	40 (14)	44 (20)	45 (12)	0.06
Years Quit*
Median	6 (8)	4 (6)	4 (11)	10 (5)	6 (7)	0.19
Prior malignancy (excluding BCC or SCC)	8 (4.2%)	3 (7.0%)	0 (0.0%)	1 (2.0%)	4 (7.1%)	0.21
Chronic pulmonary disease
Chronic bronchitis/emphysema	41 (22%)	15 (35%)	7 (16%)	9 (18%)	10 (18%)	0.15
Pulmonary fibrosis	2 (1.0%)	0 (0.0%)	0 (0.0%)	2 (4.1%)	0 (0.0%)	0.16
Exposures
Silica	6 (3.1%)	1 (2.3%)	3 (7.0%)	1 (2.0%)	1 (1.8%)	0.61
Asbestos	31 (16%)	7 (16%)	5 (12%)	13 (27%)	6 (11%)	0.12
Family history of lung cancer	47 (25%)	10 (23%)	13 (30%)	13 (27%)	11 (20%)	0.66
Insurance
Employer-based commercial	68 (36%)	14 (33%)	15 (35%)	17 (35%)	22 (39%)	0.86
Non-employer-based commercial	22 (12%)	5 (12%)	6 (14%)	7 (14%)	4 (7.1%)
Medicare	42 (22%)	11 (26%)	7 (16%)	11 (22%)	13 (23%)
Medicaid/other state program	27 (14%)	8 (19%)	6 (14%)	8 (16%)	5 (8.9%)
TRICARE/VA/Military Alaska	12 (6.3%)	1 (2.3%)	5 (12%)	2 (4.1%)	4 (7.1%)
Native/Indian/Tribal Health Services	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
Other	1 (0.5%)	0 (0.0%)	0 (0.0%)	1 (2.0%)	0 (0.0%)
None	17 (8.9%)	4 (9.3%)	3 (7.0%)	3 (6.1%)	7(13%)
Income (%)
<$25,000	61 (32%)	17 (40%)	16 (37%)	15 (31%)	13 (23%)	0.75
$25,000-$49,999	52 (27%)	11 (26%)	8 (19%)	16 (33%)	17 (30%)
$50,000-$74,999	37 (19%)	7 (16%)	9 (21%)	7 (14%)	14 (25%)
$75,000-$99,999	22 (12%)	2 (4.7%)	6 (14%)	7 (14%)	7(13%)
$100,000-$149,999	13 (6.8%)	4 (9.3%)	3 (7.0%)	4 (8.2%)	2 (3.6%)
$150,000-$199,999	1 (0.5%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	1 (1.8%)
≥$200,000	0 (0.0%)	2 (4.7%)	1 (2.3%)	0 (0.0%)	2 (3.6%)
Decline to answer	5 (2.6%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)

^*Calculated among former smokers (n=92)

Mechanical Turk (MTurk)

Body-mass index (BMI)

General Education Development (GED)

Basal cell carcinoma (BCC)

Squamous cell carcinoma (SCC)

Veterans’ Affairs (VA)

Overall, the mean self-reported likelihood of screening score was 2.7 (SD ± 1.1) with 85 participants (45%) being “very likely” or “likely” to screen. The high construal and promotion image group had the highest self-reported likelihood of screening (the lowest mean score [2.2 ± 1.0, p=0.04]) (Table 2), with 60% reporting being “likely” or “very likely” to undergo screening, whereas the high construal and prevention-focused image group had the lowest likelihood of screening (mean score of 2.9 ± 1.0), with only 30% reporting being likely to undergo screening.

Table 2.

Likelihood of Screening by Image Type

Image type	High construal, Promotion focus (n=43)	High construal, Prevention focus (n=43)	Low construal, Promotion focus (n=49)	Low construal, Prevention focus (n=56)	p-Value
Mean score: likelihood of screening* (SD)	2.2 (1.0)	2.9 (1.0)	2.7 (1.1)	2.8 (1.2)	0.04

Standard deviation (SD)

^*1=very likely to screen; 5=very unlikely to screen

We tested whether alignment between an image’s characteristics and the traits of the individual was associated with an increased likelihood of screening. We described the prevalence of personality traits among study participants (Figure 2). High construal with a promotion focus was the most prevalent personality trait combination in 77 participants (40%, 95% CI 33-48%). Low construal with a prevention focus was the least prevalent trait combination, seen in only 33 participants (17%, 95% CI 12-23%). Forty-four respondents (23%) viewed an image with characteristics that were congruent with their own personality traits. Congruence between respondent personality traits and an image’s characteristics was not associated with a greater likelihood of screening when compared to personality trait-image incongruence (48% vs 44%, p=0.62) (Table 3).

Figure 2. Personality Traits of Lung Cancer Screen-Eligible Participants.

Participants’ personality traits as assessed by the Behavioral Identification Form and Regulatory Focus Questionnaire. Error bars represent 95% confidence intervals.

Table 3.

Likelihood of Screening and Individual Personality Trait-Image Congruence

	Personality Trait-Image Congruence (n=44) % (95% CI)	Personality Trait-Image Incongruence (n=147) % (95% CI)	p-Value
Likely to screen	48 (33-63)	44 (35-52)	0.62

Confidence interval (CI)

Comment

As predicted by one psychological theory, we found that images with a high construal and promotion focus led to a higher self-reported likelihood of screening among individuals at high risk for lung cancer. However, we found no evidence that pairing of image characteristics with individual traits is associated with self-reported likelihood to screen.

Previous studies have shown that the construal level and regulatory focus of images influence behavior, however, these theories have not been well-studied in a healthcare setting involving older, high-risk subpopulations and it is unknown whether they apply in this context. These theories have previously only been examined in psychology research, where participants are generally young and healthy individuals. Additionally, numerous differences may exist between the external factors that play a role in traditional psychology research and the healthcare setting. It has been shown that awareness of one’s risk of developing lung cancer²² and smoking-related stigma^23,24 are prevalent and psychologically influential variables among high-risk smokers²⁵. These external factors could eliminate the effect of applying construal and regulatory focus theories in clinical practice. However, we provide evidence that features of an advertisement image grounded in psychological theory can influence health-related behavior in the actual clinical setting.

Although psychological theories suggest alignment of image characteristics and individual traits improves the effectiveness of an image on decision-making, it is unknown if health-related messages benefit from such alignment. Evidence of such a relationship may help health systems develop personalized communications. However, doing so requires a substantial amount of resources and patient cooperation to be successful, and for this reason it is important to know if application of psychological theories hold in the actual clinical setting. We found no evidence that personalizing messages based on construal and regulatory focus would be likely to result in more lung cancer screening. There are several potential reasons why personality trait-image congruence might not have been associated with willingness to screen in our study. One reason may be differences between participants in the psychology studies that founded these theories and our high-risk population. Additionally, qualitative studies suggest that the interaction of several psychological variables, such as anxiety and fatalistic beliefs about lung cancer, can influence behavior^26,27. Consequently, another explanation may be that these interactions reduced the impact of personality trait-image congruence on willingness to screen. Despite our null results for our second study aim, the findings are beneficial to health systems. At least as it pertains to lung cancer screening, rather than expending resources to profile the psychological traits of a population, resources can instead be directed toward developing high construal, promotion-focused images that promote lung cancer screening.

Our study has several limitations. The MTurk population may not be representative of individuals at high-risk for lung cancer. Although there is no standard by which we can readily verify our assumption of generalizability, we previously compared this population to those included in the large pragmatic National Lung Screening Trial and found that the MTurk population had a greater proportion of women, current smokers, those with at least a high school education, American Indian or Alaska Natives, and those with a history of asbestos exposure. Our findings may not be generalizable to non-white individuals or those with lower education achievement. These individuals may be less likely to pursue screening but are more likely to potentially benefit more from it. Despite our survey’s 79% response rate, there may also be bias arising from the 21% who did not respond to our survey invitation. While we do not have information on the psychological traits of non-responders, we do know responders and non-responders differ in terms of COPD rates, income, and sex. If the observed differences are not due to chance, there are potentially important differences between responders and non-responders, if income, sex, and COPD affect likelihood of screening. While self-reported likelihood to screen appears similar across income groups, they may differ in important ways across groups defined by sex and COPD status (Supplemental Table 2). These differences were not statistically significant based on confidence interval inspection, but they may be clinically important if not due to chance. This study was not powered to assess for such differences. Another limitation is that participants’ stated preference to participate in screening may differ from their revealed preference in a clinical setting. However, we are aware that even in the clinical setting, stated preferences may differ from preferences revealed through shared decision-making conversations. We could not evaluate whether an increase in self-reported likelihood to screen actually leads to higher rates of shared decision making and screening, and our Likert scale measurement of likelihood to screen has not been previously validated. Additionally, the accuracy of respondent-provided information cannot be confirmed; we are relying on self-reported sociodemographic and clinical risk factors. It is also important to note that the U.S. is considered a highly promotion-focused society. These results may not hold across cultures or societies that are believed to be primarily prevention-focused. Lastly, the images obtained from our internet search vary in numerous ways besides their construal level and regulatory focus (e.g., fonts, design, colors). We did not control for these using an A/B testing framework as we sought to evaluate currently used screening advertisements. We believe this approach is pragmatic, particularly if health systems wish to use existing images.

Our findings suggest that health systems can benefit from designing high construal, promotion-focused images as a means of steering high-risk individuals towards lung cancer screening services. Currently, health systems are unlikely to benefit from investing resources to profile high-risk individuals so that personalized messages can be developed to promote lung cancer screening. Future research can test the effect of various high construal, promotion-focused images on actual rates of shared-decision making and screening.