A rank ordering and analysis of four cognitive-behavioral stress-management competencies suggests that proactive stress management is especially valuable

Abstract

The main objective of this study was to determine the relative value of four cognitive-behavioral competencies that have been shown in empirical studies to be associated with effective stress management. Based on a review of relevant psychological literature, we named the competencies as follows: Manages or Reduces Sources of Stress, Manages Thoughts, Plans and Prevents, and Practices Relaxation Techniques. We measured their relative value by examining data obtained from a diverse convenience sample of 18,895 English-speaking participants in 125 countries (65.0% from the U.S. and Canada) who completed a new inventory of stress-management competencies. We assessed their relative value by employing a concurrent study design, which also allowed us to assess the validity of the new instrument. Regression analyses were used to rank order the four competencies according to how well they predicted desirable outcomes. Both regression and factor analyses pointed to the importance of proactive stress-management practices over reactive methods, but we note that the correlational design of our study has no implications for the possible causal effects of these methods. Questionnaire scores were strongly associated with self-reported happiness and also significantly associated with personal success, professional success, and general level of stress. Data were collected between 2007 and 2022, but we found no effect for time. The study supports the value of stress-management training, and it also suggests that moderate levels of stress may not be as beneficial as previously thought.

Introduction

The main objectives of the present study were to (a) introduce and evaluate a new instrument—the Epstein Stress Management Inventory for individuals (ESMI-i)—for assessing four cognitive-behavioral competencies that have been shown in empirical studies to be associated with effective stress management, (b) analyze data from a large, international group of English-speaking people who completed the new questionnaire online, (c) compare the relative value of the four competencies, and (d) analyze the data from a demographic perspective. By using a competencies approach, we are providing both the public and practitioners with a practical tool for measuring and potentially improving practices associated with effective stress management.

The ESMI-i joins a large number of test instruments and tools that have been developed since the 1950s to help people deal with various kinds of mental health challenges. It is also fairly unique in some respects. For one thing, it was designed for the general population, rather than for a particular group, and it was designed to measure broad competency areas rather than skills that might be helpful mainly in specific contexts. Similar inventories have been designed, for example, to assist individuals who regularly face stress in their work environments^1–3. Other inventories have been designed to measure people’s coping skills in response to specific stressors^4,5.

The ESMI-i is also available online, non-commercial, and free of charge, maintained by a nonprofit organization. Validated instruments measuring the “level” of stress people feel, such as the Holmes and Rahe Stress Scale⁶ and the Perceived Stress Scale⁷, are currently available online, and so are numerous non-validated tests of this sort, accessible at websites such as OkCupid.com. Validated inventories that assess “coping styles” also exist, such as the Multidimensional Coping Inventory⁸. These instruments differ from the ESMI-i in that they are designed to classify people, while we are careful to avoid labeling those who complete the ESMI-i; we will explore this matter later in more detail.

Other validated inventories measure cognitive-behavioral skills and overlap to some extent with the ESMI-i. However, most of these instruments were developed with cognitive-behavioral therapy (CBT) in mind and thus may not be ideal measures of stress-management competencies per se. In addition, many of these inventories, such as the Cognitive-Behavioral Therapy Skills Questionnaire⁹, the Skills of Cognitive Therapy measure¹⁰, and the CBT Skills Checklist¹¹, include items that sometimes conflate skills with reductions in symptoms of depression¹².

Validated inventories that measure stress-management or coping skills exist, such as the Proactive Coping Inventory^13,14 , Chronic Pain Coping Inventory^15–19, the Coping Inventory for Stressful Situations^20–22, the COPE Inventory^23–25 , the Coping with Stress Scale²⁶ , the Coping Intelligence Questionnaire ²⁷, the Dispositional Resilience Scale^28,29, the Stress Mindset Scale^30–32, and the Performance of Cognitive Therapy Strategies measure³³. However, they are either not available online or, in some cases, they can only be administered by licensed professionals or trained observers¹². Because many people are now relying on the internet as a major resource for self-evaluation^34–37, we believe that it is important to make validated tests widely available online. Self-help books that teach similar stress-management techniques are available to members of the general public, such as Mind Over Mood³⁸; however, such books are not free, and they take much longer to read than it takes to complete an online inventory.

By stress, we are referring to internal, usually unpleasant physiological and psychological states that are often induced by perceived environmental threats or environmental demands, which are sometimes called “stressors”^39–43. Actual environmental threats do not necessarily produce stress reactions, and the same stressor can cause different stress reactions in different people—or even no stress reaction at all^40,41.

We are not concerned in the present paper with the definitional ambiguities in the terms “stress” and “stressor.” Rather, we are focusing on stress-management practices—thoughts and behaviors that reduce stress—and we are especially interested in practices of this sort that can be both measured and trained by coaches, therapists, or counselors. The instrument we developed focuses on four classes of such behaviors; we define each class of behaviors to be a stress-management competency. The term “coping skills” is sometimes used to describe practices of this sort⁴⁴. For purposes of the present discussion, we will avoid using that term, as well as the related term “coping strategies,”⁴⁵, because we view these terms as referring mainly to reactive practices. In the present study, we will be measuring both reactive and proactive competencies, and we will use our data to compare the relative value of each type.

We believe that it is important to identify and measure stress-management competencies—especially those that can be trained—because of their enormous practical value. Stress-management competencies not only reduce levels of reported stress but have also been associated with increased functioning and well-being, as well as with improvements in mental, emotional, and physical health^46–64. Unmanaged stress is costly, both in personal and economic terms^60,65–70. Fortunately, stress management can be trained, and benefits of such training have been demonstrated^46,71–77. Levels of stress can also be measured^78–83, and so can stress-management proficiency^{17,18,21,24,27,28,30,84–87}.

The nature and value of a competencies approach

Many, if not most, test instruments used by psychologists are based on theories, and those theories are often about hypothetical constructs such as intelligence or personality traits. The methodology for developing and evaluating such instruments is quite advanced. Factor and item analyses are often employed, for example, to remove items that do not improve the statistical validity of the construct measures.

The ESMI-i is not a theory-based questionnaire, and it does not introduce or attempt to measure constructs. The ESMI-i is a competency test, developed in the spirit of a testing approach strongly advocated by David McClelland, notably in a seminal paper published in The American Psychologist in 1973⁸⁸; other experts have also been strong advocates of this approach over the years^89,90. A competencies approach to understanding and improving human performance is widely used in multiple fields and arenas—by the military⁹¹ , in healthcare⁹², in business^93,94 , in education⁹⁵, and in other areas in which human performance is important^96–101 .

As McClelland and others have pointed out, a competencies approach to understanding human functioning has some practical advantages over more traditional psychological approaches. Before he and his colleagues applied this approach to the study of leadership, for example, both businesses and armies had long been searching for “natural born leaders,” and such people exist, of course¹⁰² . But leadership, along with many other areas of human functioning—even intelligence⁸⁸—can be broken down into a number of skill areas that not only can be observed and measured; they can often be trained. Those skill areas, such as the set of skills one needs to drive a car, are not hypothetical, and they are also not constructs. They are sets of behaviors, many of which are observable. Tests that measure traits or constructs often leave people with labels, such as the trait measures yielded by the ubiquitous Minnesota Multiphasic Personality Inventory, and labels can be both demoralizing^103,104 and self-fulfilling^105–108. In contrast, competency scores simply tell people where they stand at the moment; they are often used in combination with training programs that employ questionnaires to measure post-training improvements.

Development of the ESMI-i questionnaire

The ESMI-i measures “competencies,” a term that is typically defined as “the knowledge, skills, abilities, and behaviors that contribute to individual and organizational performance”¹⁰⁹—specifically those competencies that are mainly cognitive or behavioral in nature and that might help people to reduce, eliminate, or avoid stress. Beginning in 2001, the first author of this paper, with the help of his students, set about searching the psychology literature to find peer-reviewed papers that identified skills, behavior, or knowledge that were associated with successful stress management, the goal being to use these papers to develop a test instrument that could measure the strength of such competencies. We were especially interested in competencies of the sort that therapists or counselors might be likely to teach—those, as opposed to stress-management techniques that might be taught by medical personnel, nutritionists, or other experts.

Over time, we developed a questionnaire that measured four relatively distinct cognitive-behavioral competencies: Manages or Reduces Sources of Stress, Manages Thoughts, Plans and Prevents, and Practices Relaxation Techniques. Table 1 shows the competencies, definitions, the scored items for each competency, and a list of relevant references. Below are examples of how studies published between 2001 and 2007 were employed to develop the four competency categories and to compose a total of 24 scored items. Table 1 also includes relevant references found after the questionnaire was posted online in May of 2007.

Table 1.

Four empirically-supported stress-management competencies.

Manages or Reduces Sources of Stress: You routinely manage and reduce sources of stress in life Proactive or reactive: Either Corresponding items:  “I’m comfortable seeking help from other people.”  “I have trouble keeping my work area organized.” (reverse scored)  “I consistently put important tasks ahead of unimportant tasks.”  “I have adequate shelf, file, and drawer space to serve my needs.”  “I have trouble delegating.” (reverse scored)  “I try to schedule appointments and meetings so that they won’t overlap.” References: 113,115,177,180,181

Manages Thoughts: You practice thought-management techniques in order to reduce the likelihood of perceiving an event negatively Proactive or reactive: Either Corresponding items:  “I often reinterpret events in order to lower my stress.”  “I ignore hostile people.”  “Negative events can always be reinterpreted so that they seem more positive.”  “I regularly examine and try to correct any irrational beliefs I might have.”  “My thinking is as clear and as rational as it can possibly be.” (reverse scored)  “I often try to use humor to diffuse tension.” References: 30,49,54,55,64,69,118–121,123,124,129,130,182–186

Plans and Prevents: You take a proactive approach to stress management, plan day and life and avoid destructive ways of dealing with stress Proactive or reactive: Proactive Corresponding items:  “I keep an up-to-date list of things I’m supposed to do.”  “I have a clear picture of how I’d like my life to proceed.”  “I try to fight stress before it starts.”  “I try to avoid destructive ways of dealing with stress.”  “I spend a few moments each morning planning my day.”  “I’m very vulnerable to stress when it hits.” (reverse scored) References: 22,59,67,73,118,119,121,123,129,130,185,187

Practices Relaxation Techniques: You regularly practice a variety of relaxation techniques to prevent or relieve stress Proactive or reactive: Either Corresponding items:  “I frequently use special breathing techniques to help me relax.”  “I frequently visualize soothing scenes in order to relax.”  “I regularly tense and relax my muscles as a way of fighting stress.”  “I sit and stand in special ways to help me stay relaxed.”  “Breathing is a very hard thing to control.” (reverse scored)  “I schedule relaxation time every day.” References: 51,75,132,134,135,188–193

When possible, we tried to create items that “pinpoint” specific behaviors. Items that pinpoint behavior are good predictors of actual behavior^94,110–112. So instead of saying, “I’m great at making people laugh,” we say, “I often try to use humor to diffuse tension.” That wording tells us about behavior and also, to some extent, about the frequency of that behavior. We were not able to do this for every item, but we used pinpointing as a standard for item composition.

Competency 1: manages or reduces sources of stress

In a year-long study with 100 adult residents of the Alameda County area of California, Folkman and Lazarus¹¹³ interviewed participants once every four weeks to determine what strategies they employed to help them cope with the stress they experienced as a result of activities of daily living. The researchers were guided by Richard Lazarus’ cognitive-phenomenological approach to analyzing psychological stress¹¹⁴. One of two types of coping strategies analyzed in the 1980 study was labeled “problem-focused” (p. 223), and it suggested, along with other studies (see below), that one robust category of stress management was managing sources of stress. In defining this strategy, the authors spoke of “management or alteration of the person-environment relationship that is the source of stress” (p. 223), “cognitive problem-solving efforts and behavioral strategies for altering or managing the source of the problem” (p. 224), and other actions that reduced or eliminated sources of stress.

We also used language from the Folkman and Lazarus¹¹³ study to help us construct one of our questionnaire items. In elaborating on problem-focused coping strategies, the authors stated that these strategies “include seeking information, trying to get help, inhibiting action, and taking direct action” (p. 229). Item 2 on the ESMI-i is, “I’m comfortable seeking help from other people” (Table 1).

Problem-focused coping strategies were also analyzed in a 2006 study with 67 families of young children with disabilities. Stoneman and Gavidia-Payne¹¹⁵ found that marital harmony was higher when fathers in these families employed problem-focused strategies to overcome challenges, thus reducing sources of stress.

Competency 2: manages thoughts

Thought management—often taught by counselors and therapists as part of therapeutic treatment—can be a powerful means for reducing or eliminating stress. “Reframing,” a technique most often associated with Albert Ellis¹¹⁶ and a main component of rational emotive behavior therapy¹¹⁷, is just one example of a thought-management technique. Others include cognitive restructuring^118–121, cognitive reappraisal^47,54,64, cognitive redefinition²⁶, and cognitive defusion^122–124. Most mindfulness techniques, such as acceptance and commitment therapy, incorporate methods for managing thoughts¹²⁵. Cognitive-behavioral therapy, developed by Aaron Beck in the 1960s^126,127, also emphasizes techniques aimed at reducing automatic thoughts and cognitive distortions¹²⁸. Generally speaking, people have little control over the undesirable things that happen to them, but, in theory, they have—or could be trained to have—complete control over how they interpret such events; hence, the logic of using thoughts to manage stress.

Murphy¹²⁰ conducted a meta-analysis of 64 studies that examined ways in which people managed workplace stress. The most common methods used to manage stress in work settings were meditation, muscle relaxation, cognitive-behavioral skills, and biofeedback. Cognitive-behavioral techniques, which included “thought restructuring” and other methods, proved to be especially effective in reducing what the author called “psychologic” stress (as opposed to “physiologic” stress). The Murphy study proved to be especially helpful in generating possible items for inclusion in the ESMI-i. Items based in part or in full on content from the Murphy study include, “I often reinterpret events in order to lower my stress” (item 5), “Negative events can always be reinterpreted so that they seem more positive” (item 9), and “I regularly examine and try to correct any irrational beliefs I might have” (item 10).

A comprehensive literature review by Giga et al.¹²⁹ also proved helpful in developing our Manages Thoughts competency category (as well as the competency that follows—see below). Focusing again on the work environment, the authors found that thought restructuring, reframing, and similar techniques were helpful in managing stress. Language from the Giga et al. ¹²⁹ study was helpful in developing items 5 and 9 (see above paragraph).

An experimental study by Keogh et al.¹¹⁹ also helped us develop this competency category. 209 students in the UK were randomly assigned to either a cognitive-behaviorally oriented treatment group or a no-treatment control group. Stress reduction was significantly higher in the treatment group, with students being taught, among other things, to replace irrational beliefs (such as “I am bad at taking tests” [p. 342]) with more rational ones (such as “I can take tests, if I prepare appropriately” [p. 342]). This study helped us to develop items such as the reverse-scored item, “My thinking is as clear and as rational as it can possibly be” (item 11), as well as item 10 (noted above).

Competency 3: plans and prevents

Once again, the Giga et al.¹²⁹ literature review was helpful in developing this category. It spoke specifically about the value of “plan[ning] to prevent and manage stress,” and it helped us develop two items: “I have a clear picture of how I’d like my life to proceed” (item 15) and “I try to fight stress before it starts” (item 17).

McWilliams et al.²² studied a group of 298 outpatients with major depressive disorder, having them complete multiple questionnaires, such as the Coping Inventory for Stressful Situations²¹. They concluded that planning and scheduling, among other strategies, were associated with lower levels of psychological stress. Content from this article helped us compose two ESMI-i items: “I keep an up-to-date list of things I’m supposed to do” (item 7), and “I spend a few moments each morning planning my day” (item 22). In addition, the Folkman et al. study¹³⁰, mentioned earlier, helped us compose item 18: “I try to avoid destructive ways of dealing with stress.”

Competency 4: practices relaxation techniques

The value of practicing various relaxation techniques in managing stress began to be established even before the concept of “biological stress” was introduced in the early 1900s and before Hans Selye’s breakthrough research in the 1930s on the relationship between the stress response and disease¹³¹. Edmund Jacobson’s classic book, Progressive Relaxation¹³²—based on techniques he had been developing and studying since 1915—asserted that progressive muscle relaxation exercises had multiple benefits, including improvements in memory, attention, thinking, and emotions¹³³.

Later sources have repeatedly confirmed the value that various relaxation techniques have in stress management. For example, Pawlow and Jones¹³⁴ conducted a controlled experiment on progressive muscle relaxation with 55 undergraduate students, concluding that the experimental group benefitted in multiple ways from the relaxation exercises. Among other benefits, the exercises “produced significantly reduced self ratings of perceived stress and state anxiety [and] significantly increased ratings of relaxation from immediately before to immediately after the training” (p. 381). In contrast, quiet sitting (practiced by the control group) produced no such benefits. The Pawlow and Jones¹³⁴ study helped us to compose the questionnaire item, “I regularly tense and relax my muscles as a way of fighting stress” (item 13).

Smith et al.¹³⁵ compared progressive muscle relaxation to yoga with 131 adults from South Australia in a randomized study, concluding that yoga (which included breathing exercises and postures) was as effective as muscle relaxation in producing positive outcomes, including reductions in stress and anxiety. This study helped us develop three ESMI-i items: “I frequently use special breathing techniques to help me relax” (item 1), “I regularly tense and relax my muscles as a way of fighting stress” (item 13), and the reverse-scored item, “Breathing is a very hard thing to control” (item 20).

Dummy items and internal consistency score (ICS)

As is common in competency questionnaires designed by the first author^96,98,99,101, the ESMI-i includes one dummy item for each of the four competencies assessed. Each dummy item rephrases a corresponding scored item. The purpose of having these dummy pairs is to be able, at the end of each user session, to quickly compute how closely the answers within each pair match each other. The match is computed using a modified version of Cohen’s kappa coefficient, a standard measure of inter-observer agreement¹³⁶ (see Supplementary Text S1 to compare the two formulas). We call this calculation our “internal consistency score” (ICS). In theory, if the ICS is low, we can ask a user to retake the questionnaire. In the present study, no users were asked to retake the questionnaire based on a low ICS. Instead, we elected to examine this issue as part of the data cleaning process (see below).

Methods

Participants

Before data cleaning, our dataset included 21,398 people who had completed the ESMI-i between May 3, 2007, and June 1, 2022. If someone completed the questionnaire more than once on the same day, we preserved only the first instance in which more than half the questionnaire items were answered. We also removed all cases in which self-reported English fluency was below 6 (on a scale from 1 to 10, where 10 indicated the highest level of fluency). After cleaning, 18,895 participants remained in the dataset.

The self-reported demographic characteristics of the participants were as follows (for details, see Table 2): Age ranged from 12 to 83 (M = 30.4 [SD = 14.1]). Because the ESMI-i has a Flesch–Kincaid reading level of 5.8, and because most 11-year-old children in the U.S. have completed the fifth grade, we received Institutional Review Board (IRB) approval for participants age 11 and over; however, our youngest participants (after cases were removed because of low self-reported fluency levels) were 12 (n = 19).

Table 2.

Demographic comparisons.

	n (%)	Mean manages sources of stress (SD)	Mean manages thoughts score (SD)	Mean plans and prevents score (SD)	Mean practices relaxation score (SD)	Mean total score (SD)	Significance test	P
Age
12–17	3,257 (17.2)	54.7 (18.2)	54.5 (15.0)	48.4 (18.6)	39.3 (18.8)	49.3 (12.9)	U = 19,539,532	< 0.001
18–83	15,638 (82.8)	62.7 (18.2)	57.7 (14.5)	56.3 (19.7)	42.4 (20.8)	54.8 (13.4)
Gender
Female	12,242 (64.8)	61.4 (18.2)	57.1 (14.6)	55.1 (19.6)	41.3 (20.3)	53.7 (13.3)	H = 42.7	< 0.001
Male	6,565 (34.7)	61.3 (18.7)	57.3 (14.8)	54.8 (20.0)	43.2 (20.8)	54.1 (13.7)
Other	88 (0.5)	45.8 (18.8)	57.7 (17.8)	38.0 (21.7)	34.8 (22.7)	44.1 (15.4)
Race/Ethnicity
American Indian	128 (0.7)	60.7 (18.5)	55.8 (15.9)	56.5 (21.2)	45.1 (21.7)	54.5 (14.8)	H = 330.1	< 0.001
Asian	2,379 (12.7)	64.5 (18.1)	60.8 (13.6)	58.1 (19.1)	49.7 (20.5)	58.3 (13.4)
Black	1,165 (6.2)	61.3 (19.3)	57.2 (15.3)	57.5 (20.3)	42.4 (20.4)	54.6 (14.0)
Hispanic	1,107 (5.9)	59.6 (19.3)	57.6 (15.0)	52.8 (19.4)	41.9 (20.3)	53.0 (13.8)
Other	869 (4.6)	59.8 (18.5)	58.5 (15.2)	55.0 (20.3)	42.5 (21.5)	53.9 (14.1)
White	13,097 (69.3)	60.9 (18.2)	56.4 (14.6)	54.2 (19.7)	40.3 (20.0)	53.0 (13.2)
Level of education
None	2,115 (11.2)	54.3 (18.6)	54.6 (15.4)	47.6 (18.6)	38.6 (19.1)	48.8 (13.1)	H = 604.3	< 0.001
High school	5,857 (31.0)	59.6 (18.4)	56.7 (14.6)	52.7 (19.4)	40.0 (19.4)	52.3 (12.9)
Associates	1,844 (9.8)	64.9 (17.8)	56.8 (14.3)	57.3 (20.0)	40.8 (19.9)	54.9 (13.1)
Bachelor’s	5,361 (28.4)	63.2 (18.1)	57.9 (14.6)	56.3 (19.7)	42.9 (21.0)	55.1 (13.6)
Master’s	2,952 (15.6)	64.0 (17.6)	58.9 (14.1)	59.6 (18.9)	45.9 (21.2)	57.1 (13.2)
Doctorate	644 (3.4)	62.1 (19.1)	57.5 (15.0)	59.8 (20.7)	45.9 (24.0)	56.3 (15.1)
Sexual orientation
Straight	16,713 (88.5)	62.1 (18.2)	57.2 (14.6)	55.9 (19.5)	42.1 (20.4)	54.3 (13.3)	H = 203.0	< 0.001
Bisexual	1,211 (6.4)	52.6 (19.0)	57.3 (15.6)	46.2 (20.5)	39.9 (20.4)	49.0 (14.0)
Gay or lesbian	558 (3.0)	57.0 (19.6)	56.6 (15.3)	47.6 (20.0)	39.0 (20.8)	50.0 (14.2)
Other	25 (0.1)	50.8 (19.5)	60.7 (14.9)	51.8 (23.1)	45.5 (23.0)	52.2 (17.1)
Country of origin
U.S. & Canada	12,279 (65.0)	61.2 (18.2)	56.9 (14.7)	54.9 (19.3)	41.3 (20.0)	53.6 (13.2)	U = 33,681,800	< 0.001
Other	5,712 (30.2)	61.7 (18.7)	57.9 (14.5)	55.1 (20.4)	43.4 (21.4)	54.5 (14.0)
Ever been married
Yes	7,124 (37.7)	63.6 (18.0)	56.6 (14.6)	57.7 (19.9)	42.3 (21.3)	55.0 (13.6)	U = 37,980,528.5	< 0.001
No	11,634 (61.6)	59.8 (18.6)	57.5 (14.7)	53.2 (19.5)	41.6 (20.0)	53.0 (13.3)
Ever been divorced
Yes	2,719 (14.4)	63.4 (18.1)	57.3 (14.7)	57.5 (20.2)	43.5 (22.0)	55.4 (14.0)	U = 20,128,336.5	< 0.001
No	15,992 (84.6)	60.9 (18.5)	57.2 (14.7)	54.5 (19.6)	41.6 (20.2)	53.6 (13.4)

Significance tests and p values compare mean total scores.

After cleaning by English fluency and duplicate cases, we had no need to remove cases because of low ICSs. We made this determination based on the value of Cronbach’s alpha for groups of people with differing ICSs. The group of people with ICSs between 0.4 and 0.5 (or, more precisely, 0.4 < ICS ≤ 0.5) had an alpha of 0.71, and with each successive group of people with higher ICSs (0.5–0.6, 0.6–0.7, 0.7–0.8, 0.8–0.9, 0.9–1.0), alpha increased (range 0.71 to 0.88). Because alphas greater than 0.7 are normally considered acceptable in test development^137–139 , using this criterion, we could not justify removing cases based on low ICSs. A small number of people (40 in total, 0.21% of the total N) had ICSs less than 0.4, but that was too few people for us to compute an alpha. Because we had no objective reason to eliminate these people from our study, we took the conservative course of action and let them remain (see Supplementary Table S1 and Supplementary Figure S1 for details).

Overall, 12,242 (64.8%) of our participants identified themselves as female, 6,565 (34.7%) as male, and 88 (0.5%) as other. Racial and ethnic background was as follows: 128 (0.7%) of our participants identified themselves as American Indian, 2,379 (12.6%) as Asian, 1,165 (6.2%) as Black, 1,107 (5.9%) as Hispanic, 13,097 (69.3%) as White, and 869 (4.6%) as Other; 150 individuals (0.8%) did not answer this question. Overall, 29.9% of the individuals in the sample identified themselves as non-White.

Regarding level of education completed: 2,115 (11.2%) reported not having a high school degree; 5,857 (31.0%) reported completing high school; 1,844 (9.8%) reported having an associates degree; 5,361 (28.4%) reported having completed college; 2,952 (15.6%) reported having a master’s degree; 644 (3.4%) reported having a doctoral degree; and 122 (0.6%) did not answer the question. Regarding sexual orientation: 16,713 (88.5%) identified themselves as straight; 558 (3.0%) as gay or lesbian; 1,211 (6.4%) as bisexual; 25 (0.1%) as other, and 388 (2.1%) did not answer this question. Regarding country of origin: 12,279 (65.0%) were from the United States and Canada; 5,712 (30.2%) were from 123 other countries; and 904 (4.8%) did not answer this question.

Study design

The present investigation utilized a “concurrent study design” that used criterion validity evidence, consistent with guidelines in the most recent edition of Standards for Educational and Psychological Testing¹⁴⁰, prepared jointly by the American Educational Research Association, the American Psychological Association, and the National Council on Measurement in Education. Specifically, we sought to measure the strength of the relationships between our questionnaire scores and the scores on our self-reported criterion questions. This design is called “concurrent” because we obtained questionnaire scores and criterion measures at the same time, a strategy that avoids possible temporal confounds. Results from studies employing this design are considered especially robust when the pattern of relationships between questionnaire scores and criterion measures proves to be consistent across different demographic groups.

Measures

As noted above, the questionnaire employed in the study measured four cognitive-behavioral competencies. A total score was calculated, with higher scores indicating greater stress-management competence. Separate scores were also calculated for each of the competencies. We report all questionnaire scores as a percentage of possible maximum scores rather than as raw scores. We also calculated scores for each of the four criterion questions (see Procedure below).

Procedure

Participants were first presented with brief instructions informing them, for example, that there are no right or wrong answers to the items on the questionnaire. They were then asked some basic demographic questions, following which they were asked four criterion questions regarding desirable outcomes that are sometimes associated with successful stress management, namely: How happy and fulfilled are you? How much success have you had in your personal life? How much success have you had in your professional life? How stressed do you generally feel? Answers were given on a 10-point Likert scale from Low to High (see Supplementary Figures S2-S4 for the demographic questions and the questionnaire items). After completing the 28-item questionnaire (24 items were scored), participants were given their overall questionnaire score as well as scores on the four subscales and detailed explanations about the nature of each competency (see Supplementary Fig. S5 for a screenshot of the results page). Primary access to the questionnaire was at the URL https://MyStressManagementSkills.com. Over time, links to the questionnaire appeared elsewhere on the internet, a process over which we had no control. We also had no control over the demographic characteristics of the participants (see Discussion).

Ethics statement

The federally registered Institutional Review Board (IRB) of the sponsoring institution (American Institute for Behavioral Research and Technology) approved this study with exempt status and a waiver of the requirement for informed consent under U.S. Department of Health and Human Services regulations (45 CFR 46.116(d), 45 CFR 46.117(c)(2), and 45 CFR 46.111) because (a) the anonymity of participants was preserved and (b) the risk to participants was minimal. The IRB is registered with the Office for Human Research Protections under number IRB00009303, and the Federalwide Assurance number for the IRB is FWA00021545.

Results

Regressions and factor analysis

Linear regression was used to determine which competencies were most strongly associated with self-reported levels of happiness, personal success, professional success, and general level of stress. Notably, the Plans and Prevents competency proved to be the best predictor of all four criterion variables (Tables 3 and 4). Please note that by using the language of prediction, we do not mean to imply causation.

Table 3.

Regression results showing how well the most predictive competency predicted answers to the criterion questions.

Competency	Happiness			Level of stress			Personal success			Professional success
	Standardized			Standardized			Standardized			Standardized
	β	t	r2	β	t	r2	β	t	r2	β	t	r2
Plans and Prevents	0.46***	71.42	0.21	− 0.34***	− 49.07	0.11	0.38***	56.30	0.14	0.35***	50.38	0.12

Results are shown for one-component models in stepwise regressions, one for each of the four criterion questions.

***p < 0.001.

Table 4.

Regression results showing how well the four competencies predicted answers to the criterion questions.

Competency	Happiness			Level of stress			Personal success			Professional success
	Standardized			Standardized			Standardized			Standardized
	β	t	r2	β	t	r2	β	t	r2	β	t	r2
Plans and Prevents	0.31***	35.62	0.24	− 0.21***	− 22.47	0.13	0.25***	27.58	0.17	0.22***	24.56	0.14
Manages Sources of Stress	0.19***	23.34		− 0.15***	− 17.67		0.20***	23.55		0.18***	21.59
Practices Relaxation Techniques	0.06***	7.77		− 0.07***	− 8.20		0.03***	3.46		0.02*	2.47
Manages Thoughts	0.04***	5.63		− 0.02**	− 2.71		N/A	N/A		N/A	N/A

Results are shown for multi-component models in stepwise multiple regressions, one for each of the four criterion questions.

*p < 0.05, **p < 0.01, ***p < 0.001.

An exploratory principal components factor analysis for the 24 scored items in the questionnaire yielded a Kaiser–Meyer–Olkin sampling adequacy of 0.89, which is well above the recommended cutoff of 0.6, as well as a highly significant Bartlett’s Test of Sphericity (p < 0.001). The factor analysis yielded four components that overlap our original four competencies and that can reasonably be described as: (1) Plans Ahead, (2) Practices Relaxation Techniques, (3) Regulates and Manages Stressors, and (4) Recognizes Weaknesses (Table 5). The results of the factor analysis were not used to revise the original competencies or items, because these competencies are practical skillsets, not hypothetical constructs. We did not conduct a confirmatory factor analysis for the same reason.

Table 5.

Exploratory factor analysis for the 24 scored items.^a

Component
Item	Plans ahead	Practices relaxation techniques	Regulates and manages stressors	Recognizes one’s weaknesses
7	0.74
22	0.67
25	0.59
21	0.58
4	0.53			0.41
12	0.47			0.40
13		0.74
14		0.71
3		0.64
1		0.64
27		0.42
9			0.65
10			0.65
5			0.53
18			0.50
19			0.45
8			0.40
17			0.38
2
23				0.72
24				0.68
11				0.53
20				0.50
15				0.36

Extraction Method: Principal Component Analysis.

Rotation Method: Varimax with Kaiser Normalization.

Values less than 0.35 are omitted.

^aCorrected for reverse-scored items.

Reliability and validity evidence

The questionnaire had moderate but acceptable internal-consistency reliability (Cronbach’s alpha = 0.79; Guttman split-half = 0.70)^137–139 . Because the study was conducted over the internet and because we could not collect contact information for our participants (in order to preserve their anonymity), test–retest reliability could not be assessed. We also did not develop an alternate form of the questionnaire, so alternate-form reliability could not be estimated.

Internal consistency scores for the four competencies varied considerably: Manages or Reduces Sources of Stress: Cronbach’s alpha = 0.60; Guttman split-half = 0.64. Manages Thoughts: Cronbach’s alpha = 0.34; Guttman split-half = 0.37. Plans and Prevents: Cronbach’s alpha = 0.64; Guttman split-half = 0.68. Practices Relaxation Techniques: Cronbach’s alpha = 0.65; Guttman split-half = 0.61.

Regarding validity evidence, total scores were correlated with scores on our criterion questions, as we had predicted: Total scores were positively correlated with participants’ self-reported level of happiness (Spearman’s ρ = 0.45, p < 0.001), personal success (ρ = 0.35, p < 0.001), and professional success (ρ = 0.32, p < 0.001), and negatively correlated with participants’ general level of stress (ρ = -0.33, p < 0.001). Because this was an internet-based study in which the anonymity of participants was protected, we could not assess validity by comparing scores on our questionnaire to scores on comparable questionnaires (see Discussion). (Because scores on the ESMI-i lie on an ordinal scale, nonparametric statistical tests such as Spearman’s ρ, the Mann–Whitney U, and the Kruskal–Wallis H are used throughout this study, except in our regression analyses. Nonparametric regressions are generally used only when extreme values might distort the results¹⁴¹; outliers are unlikely with an instrument like the ESMI-i in which scores are constrained.).

Although not specifically predicted, the validity of the measuring instrument is also suggested by the fact that the mean total score for those who reported having had stress-management training was significantly higher than the mean total score for those who did not (M_Yes = 58.8 [13.2], M_No = 52.6 [13.3], U = 19,888,845, p < 0.001, r = 0.18). In addition, overall questionnaire scores were positively correlated with the number of hours of stress-management training participants reported having (ρ = 0.24, p < 0.001).

Gender, race, and other demographic effects

The overall mean total score was 53.8 [13.5] and subscale means were as follows: Manages or Reduces Sources of Stress (M = 61.3 [18.4]), Manages Thoughts (M = 57.2 [14.7]), Plans and Prevents (M = 54.9 [19.7]), and Practices Relaxation Techniques (M = 41.9 [20.5]).

We found a significant but not necessarily substantial effect for education level (M_None = 48.8 [13.1], M_Highschool = 52.3 [12.9], M_Associates = 54.9 [13.1], M_Bachelors = 55.1 [13.6], M_Masters = 57.1 [13.2], M_Doctorate = 56.3 [15.1], H = 604.3, p < 0.001, E²_R = 0.03), and scores were higher for participants who reported having been married (M_Yes = 55.0 [13.6], M_No = 53.0 [13.3], U = 37,980,528.5, p < 0.001, r = 0.07) and also, surprisingly, for participants who reported having been divorced (M_Yes = 55.4 [14.0], M_No = 53.6 [13.4], U = 20,128,336.5, p < 0.001, r = 0.05).

We found a significant but not necessarily substantial effect for gender (M_female = 53.7 [13.3], M_male = 54.1 [13.7], M_other = 44.1 [15.4], H = 42.7, p < 0.001, E²_R = 0.00) and no significant male/female difference (U = 39,631,749.5, p = 0.12, r = 0.01). Participants ages 18 and older scored significantly higher than minors (M_12-17 = 49.3 [12.9], M_18-83 = 54.8 [13.4], U = 19,539,532, p < 0.001, r = 0.15). (Note that because this is a large-n study, statistical significance is not necessarily a good indicator of the importance of mean differences. For this reason, we also have included two different measures of effect size: r, where we are comparing two means, and epsilon-squared, where we are comparing three or more means ¹⁴². We also found a significant but not necessarily substantial effect for ethnicity, with respondents of Asian descent outscoring all other ethnicities (M_{AmericanIndian} = 54.5 [14.8], M_Asian = 58.3 [13.4], M_Black = 54.6 [14.0], M_Hispanic = 53.0 [13.8], M_White = 54.0 [14.1], M_Other = 53.0 [13.2], H = 330.1, p < 0.001, E²_R = 0.02; Asian vs. non-Asian: M_Asian = 58.3 [13.4], M_NonAsian = 53.2 [13.4], U = 15,133,970.5, p < 0.001, r = 0.13), a finding that is consistent with other research^143–145.

We also found a significant but not necessarily substantial effect for sexual orientation, with self-labeled straights outscoring self-labeled gays, lesbians, and bisexuals (M_Bisexual = 49.0 [14.0], M_Gay/Lesbian = 50.0 [14.2], M_Straight = 54.3 [13.3], M_Other = 52.2 [17.1], H = 203.0, p < 0.001, E²_R = 0.01; straight vs. non-straight: M_Straight = 54.3 [13.3], M_NonStraight = 49.4 [14.1], U = 11,946,435, p < 0.001, r = 0.10). Participants outside the U.S. and Canada scored higher than participants from the U.S. and Canada (M_US/Canada = 53.6 [13.2], M_Other = 54.5 [14.0], U = 33,681,800, p < 0.001, r = 0.03), possibly because of the higher scores of Asian participants in the study. We did not have enough participants in individual countries outside the U.S. and Canada for us to conduct, with sufficient statistical power, a country-by-country analysis; a larger sample might allow us to conduct such an analysis in future years. Age proved to be a small but significant predictor of questionnaire scores (ρ = 0.12, p < 0.001).

Apparent value of low stress

Our results suggest that high stress is associated with low levels of happiness, personal success, and professional success; that low stress is associated with high levels of these outcomes; and that the benefits sometimes associated with a moderate level of stress might not be as beneficial as previously thought (Figs. 1 and 2; see Discussion). Although the spikes toward the center of the curves in Fig. 1 could be interpreted as indicating possible benefits of moderate stress, when one looks closely at the happiness, personal success, and professional success ratings reported by people who experience different levels of stress in their lives, it seems evident that low stress is more consistently associated with desirable outcomes (Fig. 2).

Figure 1.

Relationship between self-reported levels of stress and self-reported levels of happiness, personal success, and professional success. Although higher stress is generally associated with poorer outcomes (note the overall downward slopes of the curves), the upward spikes in the center of the graph are sometimes mistakenly interpreted to mean that moderate stress is beneficial. Vertical bars show 95% confidence intervals.

Figure 2.

Histograms showing distributions of self-reported levels of happiness, personal success, and professional success, separated into three categories of level of stress. Although one can find high levels of happiness, personal success, and professional success in the bottom two rows of graphs, the patterns of scores are more nearly optimal in the top row, which shows data only for people reporting their overall level of stress as very low (1 on a scale from 1 to 10).

Year-by-year analysis

Because our data were collected over a period of more than 14 years, we asked whether any trends were evident in scores, as well as in demographic characteristics of the sample. Although statistically significant changes were evident in both scores and demographic characteristics over the course of the study (Table 6), we did not find a linear trend in the mean total scores (p = 0.813, r² = 0.005, β = 0.07, t = 0.24).

Table 6.

Differences in mean scores and changes in key demographic groups by year.

	2008	2009	2010	2011	2012	2013	2014	2015
n	1615	724	735	3495	1692	1193	1248	1596
Mean, total score (SD)	53.1 (13.8)	53.3 (13.4)	52.4 (13.6)	52.4 (13.0)	53.7 (13.0)	56.1 (13.0)	54.4 (12.6)	54.5 (13.6)
Mean, manages sources of stress (SD)	61.1 (18.1)	60.4 (18.0)	60.4 (17.9)	59.1 (17.8)	62.1 (17.9)	64.2 (17.5)	63.2 (18.2)	62.3 (18.7)
Mean, manages thoughts (SD)	56.2 (14.6)	58.1 (13.7)	57.0 (15.1)	56.1 (14.9)	56.5 (14.9)	58.0 (14.1)	56.8 (14.2)	57.8 (14.5)
Mean, plans and prevents (SD)	53.9 (19.9)	53.6 (19.7)	52.8 (20.0)	53.2 (18.9)	55.1 (19.0)	59.1 (18.9)	57.0 (18.8)	56.7 (19.8)
Mean, practices relaxation (SD)	41.1 (21.8)	41.1 (21.5)	39.5 (20.4)	41.3 (19.8)	41.2 (19.6)	43.4 (20.9)	40.5 (19.3)	41.3 (20.3)
Mean age (SD)	40.9 (16.7)	31.6 (12.6)	32.9 (13.0)	29.0 (15.6)	29.9 (14.0)	31.6 (13.3)	29.2 (11.9)	27.7 (12.5)
Gender (% Fem.)	61.2	67.7	68.0	57.7	64.1	69.7	72.4	65.7
Race (% White)	76.9	77.8	81.2	79.4	78.2	73.9	64.1	75.4

	2016	2017	2018	2019	2020	2021	Significance testa
n	1147	786	795	697	771	845
Mean, total score (SD)	52.1 (14.3)	55.6 (13.7)	53.3 (12.3)	53.2 (13.6)	51.2 (13.7)	54.9 (13.3)	159.5***(H)
Mean, manages sources of stress (SD)	59.1 (19.8)	63.3 (19.1)	60.5 (18.1)	59.4 (19.9)	58.0 (18.9)	61.0 (18.1)	166.3***(H)
Mean, manages thoughts (SD)	56.8 (15.1)	57.0 (14.5)	57.9 (14.1)	58.4 (14.1)	56.5 (15.3)	58.0 (14.7)	46.0***(H)
Mean, plans and prevents (SD)	52.3 (21.1)	57.6 (19.8)	54.0 (19.1)	53.3 (20.8)	50.4 (21.0)	55.6 (20.0)	196.3***(H)
Mean, practices relaxation (SD)	40.1 (20.4)	44.5 (21.0)	40.6 (18.6)	41.6 (19.5)	40.0 (20.3)	44.8 (20.7)	71.1***(H)
Mean age (SD)	28.0 (12.6)	30.8 (12.6)	27.6 (11.5)	27.5 (12.1)	31.3 (13.9)	29.5 (12.5)	90.4***(F)
Gender (% Fem.)	67.0	67.6	69.3	72.0	72.5	70.9	378.7***(X2)
Race (% White)	64.7	66.4	64.8	58.2	60.1	50.8	1115.0***(X2)

Only years in which 12 months of data were collected are shown (n = 17,339). % Female is shown for gender because the majority of participants were female. % White is shown for Race/Ethnicity because the majority of participants were White.

^aMean total and competency scores were compared using the Kruskal–Wallis H. Mean ages were compared using a one-way ANOVA. Percentages for gender and race were compared using Pearson chi-square.

***p < 0.001.

Discussion and limitations

The present study sheds light on various aspects of people’s ability to manage stress. One of its greatest limitations—that the data were collected over the internet—is also a strength. On the downside, internet sampling gives one no control over demographics, and all participants are self-selected. Our sample presumably consisted of people who were concerned about stress or how they managed it. This could mean, among other things, that our mean level of self-reported stress (6.5 out of 10) is higher than that of the general population and, perhaps, that the stress-management proficiency level we found (M = 53.8) is lower than normal. A 2013 report by the American Psychological Association⁶⁵ states that the average stress level for Americans is 4.9 out of 10, 1.6 points below the mean we found. We might also be attracting people with abnormally low levels of happiness or success.

On the upside, the internet allows researchers to look at a large, diverse, international sample, which almost certainly yields more valid findings than the proverbial pool of second-year college students^146–148. There is also accumulating evidence that people are more honest when answering personal questions through anonymous internet surveys than perhaps through any other means^149–153; a recent study by Robertson et al.¹⁵⁴ suggests that anonymous internet surveys yield more valid responses than sixteen other common survey techniques. Surveys yield especially valid responses when people are completing them voluntarily and they know that the results will not be used by supervisors or other authority Figures.^155,156 . For these reasons, we conjecture that our participants were probably honest in their responding. We also have no a priori reason to believe that the relationships we have found among variables—for example, the negative correlation between total questionnaire scores and self-reported levels of stress (ρ = − 0.33, p < 0.001)—are invalid.

We also have no reason to doubt the validity of some of the more distinctive demographic findings in the study, particularly where such findings are consistent with those of other research. Especially notable in this study is the relatively high mean score of participants identifying as Asian. Other studies looking more directly at this issue have also found that various Asian groups are better at managing stress than non-Asians, perhaps because of the collectivist nature of many Asian cultures¹⁵⁷. Tweed et al.¹⁴⁵ found, for example, that East Asian Canadians reported using internal strategies to manage stressful situations more often than European Canadians did. In collectivist cultures, people tend to be more mutually supportive than in individualistic cultures^158,159, and Asian cultures also tend to teach explicit techniques—yoga, meditation, tai chi, and so on—which have been shown to improve well-being and lower stress^71,160,161. In many non-Asian cultures, well-being is often sought through self-destructive means (alcoholism, drug abuse, overeating) or, at best, left to chance. Similarly, our findings that self-reported straights outscored self-reported non-straights on the ESMI-i and that self-reported straights reported experiencing less stress than self-reported non-straights are consistent with the findings of other researchers^162–165.

Our study found no significant difference between scores for males and scores for females. Researchers disagree about gender differences in both stress-management proficiency and perceived stress levels. Some studies suggest that women are more likely to utilize emotion-focused coping in response to stressors while men more often use problem-focused coping¹⁶⁶. The 2013 report on stress published by the American Psychological Association stated that women report higher stress levels than men⁶⁵. However, and consistent with our results, some studies have found that gender differences in coping styles are not apparent when confounding factors such as socioeconomic status and race are controlled for¹⁶⁷.

The two largest demographic effects we found should be studied in further detail in future studies. Self-labeled straights outscored non-straights by 4.9 points (r = 0.10), and adults outscored minors (ages 12–17) by 5.5 points (r = 0.15). Considerable research has examined the emotional problems often experienced by non-straights (brought about, most likely, by entrenched heteronormativity in most cultures^168–170), but why straights should score higher on a test of stress-management competencies is unclear. The age difference seems less mysterious. Competencies take time to learn, after all¹⁷¹, but it would be interesting to look at this learning process in more detail, especially over the teen years.

A second notable limitation of the present study is that it is correlational in design. In follow-ups to this study, one could, by employing either between-subjects or within-subjects experimental designs, assess the possible causal impact of each of the four competencies we have examined in this report.

Perhaps the clearest and, in some sense, the most surprising finding in this study is that proactive stress-management methods appear to be more helpful than reactive ones. All four of our criterion variables were best predicted by the Plans and Prevents competency, of which all questionnaire items describe proactive methods of fighting stress—in other words, ways of trying to ensure that stressful situations never arise (Tables 3 and 4). A planning competency also emerged in our factor analysis (Table 5). Unfortunately, our respondents scored relatively poorly on Plans and Prevents (M = 54.9), which ranked third on actual competency scores. We note that our findings about proactive methods do not necessarily show that such methods are more beneficial than reactive ones; it might simply be the case that people with lower stress levels rely less on reactive coping strategies than on proactive ones. Again, questions of cause and effect can only be answered with experimental research.

Our study also yielded intriguing findings regarding the supposed value of moderate stress. Ever since the formation of the Yerkes-Dodson law in the early 1900s, researchers have suggested that moderate levels of stress (at least for stressors of certain types) are beneficial^172–176. Our study suggests another possibility—namely, that the bulge that often appears in the center of performance or other curves where the stress level is moderate is a statistical anomaly. As we noted earlier, this seems evident when we examine the relationship between participants’ self-reported levels of stress and their self-reported levels of happiness, personal success, and professional success (Fig. 1), as well as when we look closely at the distributions of self-reported levels of happiness, personal success, and professional success when separated into low, medium, and high of levels of self-reported stress (Fig. 2).

The strong relationship (ρ = 0.45, p < 0.001, ρ² = 0.20) we found between total questionnaire scores and self-reported happiness is also notable, suggesting the importance of stress management in having a happy life¹⁷⁷ (although, once again, we remind the reader that this is a correlational study). Unfortunately, our results suggest that people are generally poor at stress management; the mean percentage score on the questionnaire was 53.8%, with Practices Relaxation Skills having a mean score of only 41.9%. Our study confirms the need to educate and train people in how to manage stress, although, as noted above, our questionnaire scores might be lower than the average scores one would find in the general population. Fortunately, although the present study employed a concurrent design¹⁴⁰, not an experimental one, our results are consistent with the view that stress-management training has value; as noted earlier, study participants who had had such training scored significantly higher than participants who had not, and questionnaire scores were positively correlated with the number of training hours reported.

We have already mentioned several ways in which our analysis was constrained because we collected our data online. These limitations are not trivial, because collecting data on the internet—especially when one is required to protect the anonymity of the participants—means that some of the standard tools used to validate new test instruments cannot be employed. Ideally, one would like to compare test scores to those obtained on previously validated tests. Again, ideally, one would like to be able to measure the stability of test scores by readministering the test to the same cohort after different periods of time have passed. One would also like to use multiple measures to validate the test, such as ratings by peers or clinicians. None of these methods is possible given our current design.

As noted above, future versions of the ESMI-i might include additional competency categories. People can mitigate stress using many tools, such as exercise or changes in diet; the present study focuses on cognitive-behavioral methods that might be taught by therapists and counselors. More detailed demographic analyses might also be conducted, especially with larger samples. The current online version of the ESMI-i (as of July 8, 2024) already includes additional gender and sexual orientation categories. The present study also assumes, implicitly, that stress management is the same in all cultures around the world, which is clearly not the case. Stressors themselves are culturally based^{163,165,178,179}, and so are effective techniques of stress management¹⁴⁵.

Author contributions

R.E. conceptualized the study, supervised all aspects of the research, and wrote the original draft. J.A., C.G., V.Z., and L.Z. assisted with statistical analysis and the literature review. V.Z. prepared Figs. 1 and 2. All authors reviewed the manuscript.

Data availability

Anonymized raw data employed in the present study will be made available at Zenodo.com upon acceptance of the manuscript. Anonymized raw data can also be requested by writing to info@aibrt.org. Anonymization was required to comply with the requirement of the sponsoring institution’s Institutional Review Board that the identities of the participants be protected in accordance with HHS Federal Regulation 45 CFR 46.101.(b)(2).

Competing interests

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-68328-4.