In the biomedical approach to health in PE, the benefits of PE have largely been linked to improving or maintaining population physical fitness and/or hygiene (3, 6). The consideration of children's varying physical abilities has contributed to a gradual shift from emphasizing physical fitness to just increasing PA. At the turn of the millennium, there was a growing public concern for physical inactivity, sedentary behavior, and the rise of noncommunicable conditions such as obesity and type II diabetes. This contributed to a call from professionals and politicians alike for more PA in schools in general (7) and in PE specifically (8, 9). In Norway, the advocates of PE's public health potential even went as far as to explicitly state that health, represented by increased energy expenditure, was the only relevant goal of PE and that this had to be prioritized before any pedagogical concerns about learning a curricula (10). Following this line of argument, the health benefits of PE have mainly been linked to the cumulative amount of PA it produces in its students (e.g. 11–16). Several scholars have critiqued the usefulness of a biomedical health perspective for the PE context (e.g. 8, 17, 18); they warned against how this might lead to adopting a healthism ideology (19, 20), which situates the problem and solutions of health and disease at the level of the individual. Such approaches could end up rewarding the athletic and those who are already engaging in regular PA and sports activities, while the less athletic and those who are not so active are put at a disadvantage. This is reflected in the Norwegian PE context, where it is the sports-active youth who seem to collect the most benefits from the subject (21). In particular, girls who do not participate in sports seem to benefit the least from PE (22). Importantly, well-intended programs for health promotion through PE that emphasize increased PA could end up being detrimental to some students’ health and well-being (18), if they are not sensitive to the various needs related to adolescents background and previous experiences. In line with this critique, it has been argued that health should be viewed as a ‘learning object’ within educative contexts. Quennerstedt (23) made a case for a resource-based health perspective, in which attention can be drawn to the abilities, knowledge, and skills that pupils should develop through education, rather than merely increased levels of PA, which characterizes a biomedical health perspective in PE. Mong and Standal (5) found that, in most such programs, teaching strategies were primarily based on instructive methods of predetermined content. On the other hand, among resource-based or critical approaches, there has been a greater tendency toward recommending non-instructive and participatory strategies, in which teachers, together with the students, become the facilitators of learning.

Shifting the focus from maximizing PA to learning for health through participatory methods may be a beneficial transition in PE. However, this shift does not necessarily address the criticism of healthism: too much responsibility for one's health is placed on the individual. Therefore, it is pivotal to balance the focus on individual skills for health with an emphasis on the collective attributes for health and well-being. In PE, such abilities can include appreciating and supporting the needs of others, as well as participating in the democratic processes and decisions that concern the collective.

CHL provides a framework for thinking about health as both an individual and collective matter (24). In line with the resource-based health perspective, the concept has been put forward and promoted to avoid the pitfalls of traditional health education efforts that are dominated by the instruction and delivery of predetermined content (25, 26). The concept has roots within emancipatory critical pedagogy (27–29) and consists of three overlapping and interconnected domains: information appraisal, understanding the social determinants of health, and the abilities that enable actions that can promote health and well-being in a collective (30, 31). The first domain of information appraisal is a complex cognitive skill that is essential for pupils to gain agency for health and well-being (32, 33); it includes abilities and strategies to judge the credibility and relevance of health-related information (31). The majority of CHL research revolves around these cognitive abilities, both in general public health research (24, 30, 34) and in school contexts (35). The second domain partially depends on the cognitive skills of appraising and understanding health-related information in its proper context, but it also borders on affective and attitudinal abilities (36). The domain involves an appreciation of how individuals have different opportunities for health and, subsequently, that there is a balance between individual and collective responsibility for health. Finally, the third domain revolves around the ability to act to the benefit of one's own and others’ health and well-being, and these actions inevitably depend on personal and social abilities (30). In a recently developed measurement framework for CHL in school contexts, these abilities are operationalized as social actions that adolescents can take to become active agents for health and well-being in a collective (31, 36).

As most of the empirical health-related research in PE tend to derive from a biomedical understanding of health (5) we aim to shed light on alternative ways in which PE can contribute to overall education about health in schools, by examining the relationship between academic achievement in PE and CHL. Although PE might be a suitable context for nurturing all aspects of CHL, we have chosen to focus on the third domain in the present study because PE is an appropriate context for developing social interaction abilities (37). The subject qualitatively differs from regular classroom settings by providing a more dynamic and interactive learning environment (38). In addition, the new national curriculum for PE highlights participation and cooperation as one of three core elements of the subject. Abilities to recognize differences, include others and reflect around equality are emphasized (1). Therefore, there are reasons to believe that CHL-C abilities can be nurtured within the PE context. Because achievement in PE are unequally distributed in various groups of students, we control for parents’ education, participation in sports club activities, and leisure PA. The following hypothesis is tested: Academic achievement in PE is positively associated with levels of CHL, and the relationship remains when controlling for parents’ education, leisure PA, and participation in sport club activities.

This cross-sectional study was part of the Literacies for Health and Life Skills project at Oslo Metropolitan University (39) and was conducted in five partner schools during the autumn of 2021. All the lower secondary schools were in medium-income municipalities in the area surrounding the capital of Norway, and they varied in size, with two small schools (< 200 pupils), one medium-sized school (250–350 pupils), and two large schools (> 450 pupils). Approximately 1,592 pupils aged 13–15 years old attended the schools at the time of data collection. After school management agreed to participate, two video presentations with standardized information for the pupils were sent to the schools. The teachers were instructed to use the first one to introduce the study before handing out written consent forms to pupils and their parents. The second provided standardized information to be presented shortly before conducting the survey, repeating the main purpose of the study, and instructing about the practicalities. In total, 522 pupils consented and responded to the digital survey during school hours. Those who did not provide written consent from their guardians were given alternative assignments.

CHL was measured with the two scales from the Critical Health Literacy for Adolescents Questionnaire (CHLA-Q) (31, 36) that specifically target the third domain of CHL. Both scales consist of three indicators measuring perceived abilities to support others (CHL-C1) and perceived abilities to participate in discussions regarding health (CHL-C2). The indicators are measured on a five-point Likert scale (1: completely incorrect; 2: incorrect; 3: sometimes correct; 4: correct; 5: completely correct), and each item starts with the phrase: “I am a person who…” (chl36: …can help others if they are not doing well; chl37: …can contribute to the well-being of others in my class; chl38: …can help find solutions that are acceptable to all parties; chl39: …can easily talk to others, even if I don't know them very well; chl40: …can share information with others about factors that influence health; chl41: …believe my knowledge about health could be useful for others). Descriptive statistics for each indicator are presented in Supplementary Materials. Performance in PE was measured using one indicator asking pupils which grades (1–6) they expected to achieve in PE this semester. Leisure PA was measured with one question: “Outside of school hours, how many times during a week do you participate in PA, to the extent that you become out of breath and/or begin to sweat?” (1: 0–1 time; 2: 2–3 times; 3: 4–5 times; 4: more than 5 times). Participation in sports club activities was measured with one item: “Are you an active member of a sports club? (e.g., football, handball, cross-country skiing, tennis, gymnastics, athletics, swimming)” (1: never been a member; 2: no, but I have been a member previously; 3: yes). The indicators of PA and SPORT were inspired by the national survey study of PA in Norway (40). Parents’ education (ParEd) was initially measured using two items: one for mothers’ education and one for fathers’ education. Both items were formulated in the same way: “mothers/fathers’ education?” (1: Higher education [university or college], 2: High school/upper secondary school; 999: don't know/doesn't want to answer). In the final data analysis, we combined the responses for the parents’ education (0: only 1 or none of parents have higher education/don't know; 1: both parents have higher education), participation in sports club activities (0: not active; 1: active member), and PA (1: > 3 times, 0: < 4 times) into two categories.

All statistical analyses were performed using RStudio (RStudio Team, version 2021.9.0.351). The “tidyverse” packages (41) was used for data preparation, and the “lavaan” package (42) was used for confirmatory factor analysis (CFA) and structural equation model (SEM) analysis. Following recommendations for SEM analysis with Likert-type items of five categories or less, all variables in our study were treated as ordinal (43, 44), and the means and variance adjusted unweighted least square (ULSMV) estimator was applied (45–47).

Initially, we examined mean group differences in gender, grades, ParEd, leisure PA, and SPORT for the outcome variables academic achievement in PE, CHL-C1, and CHL-C2. The independent sample t-test was used in cases with two groups, and an ANOVA test was used to test for statistically significant differences between grades.

A total of 521 adolescents responded to the full survey, which yielded a response rate of 33% when calculated as responses from the total eligible population. On average, it took 17 min to complete the full survey. Table 1 shows that approximately the same number of boys and girls participated. Two-thirds (total = 66.8%, boys = 70%, girls = 63%) of the sample were active members of sport clubs, and 40% reported that both parents had more than three years of higher education. Almost half the sample (total = 49%, boys = 58%, girls = 41%) reported being physically active three times or more during a week, outside of school hours. The overall mean of the expected grades in PE was 5.0, while the mean ranged from 4.5 to 5.3 in the various group variables (gender, grade, parents’ education, member of sport clubs and level of PA). For CHL-C1, the means ranged from 11.3 to 12.1, with the total mean at 11.7. The CHL-C2 sample means were slightly lower at 9.7, with group means ranging from 9.2 to 10.2. Those who were active members of sports clubs and reported being physically active four times or more during a week scored significantly higher than their peers in all outcome variables. This was also the case for pupils who reported that both parents had more than three years of higher education. Boys scored significantly higher than girls on the expected grades in PE and CHL-C2 scales; however, girls scored higher on the CHL-C1 scale, though the difference was not significant. There were no significant differences between grades.

Before performing the SEM analysis, we estimated a measurement model with the latent variables (Table 2). Covariance was estimated freely between all latent variables. The factor loadings were strong, ranging from 0.699 to 0.846 and significant (p < 0.000). The chi-square test was significant; however, the residuals of the model were small, as reflected with the uSRMR = 0.032. When adjusting for the average communality, the model also displayed a close fit (uSRMR/R¯ ²= 0.052). No individual residuals were above 0.10. Thus, we retained the model and continued with stepwise SEM analysis.

In the first step we added regression paths from PE to CHL-C1 and CHL-C2. We found a significant relationship between the predictor (PE) and outcome variables CHL-C1 and CHL-C2 (Step 1: β^_PE→_CHL−C1= 0.310, p = 0.000, β^_PE→_CHL−C2= 0.368, p = 0.000). In the second step, we added parents’ education (ParEd) to the model as a predictor of PE and both CHL scales. ParEd had a significant direct effect on PE (β^_ParEd→_PE= 0.217, p = 0.000) but not on the CHL scales; the relationship between PE and CHL became, as expected, slightly weaker, though negligibly so (Step 2: β^_PE→_CHL−C1= 0.301, p = 0.000, β^_PE→_CHL−C2= 0.353, p = 0.000). The fit indices improved as we added complexity to the model (χ²_mvadjusted= 29.646 (df = 16, p < 0.020), CFI = 0.992, RMSEA = 0.040 [0.016, 0.063], SRMR = 0.038). In step 3, we repeated the procedure from the previous step, but now, we added PA to the model as well. The relationship between PE and the CHL scales became somewhat weaker because the additional variables had their own direct paths to the CHL scales; however, the reduction was relatively small (Step 3: β^_PE→_CHL−C1= 0.268, p = 0.000, β^_PE→_CHL−C2= 0.326, p = 0.000). PA had a stronger effect on PE than ParEd (Step 3: β^_PA→_PE= 0.441, p = 0.000, β^_ParEd→_PE= 0.130, p = 0.020). None of the added variables had a significant direct effect on CHL. Model fit did not change significantly between step 2 and step 3 (step 3: χ²_mvadjusted= 36.585 (df = 20, p < 0.013), CFI = 0.990, RMSEA = 0.040 [0.018, 0.060], SRMR = 0.037). Following the same procedure in the final step, we added participation in sports clubs to the model. The relationship between the PE and CHL scales remained approximately the same (Step 4: β^_PE→_CHL−C1= 0.264, p = 0.001, β^_PE→_CHL−C2= 0.351, p = 0.000). The three background variables all had a significant and positive effect on PE (Step 4: β^_ParEd→_PE= 0.108, p = 0.045, β^_PA→_PE= 0.257, p = 0.000, β^_SPORT→_PE= 0.332, p = 0.000). None of the background variables had a significant direct effect on CHL in this model.

The final model (Figure 2) displayed an excellent model fit (χ²_mvadjusted= 31.156 (df = 24, p < 0.149), CFI = 0.996, RMSEA = 0.024 [0.000, 0.045], SRMR = 0.036). In Table 3, we present the results from the regressions in the model. We found a positive and significant effect from PE on both the dependent variables CHL-C1 (β^_PE→_CHL−C1= 0.264, p = 0.001) and CHL-C2 (β^_PE→_CHL−C2= 0.351, p < 0.000). None of the background variables had a significant direct effect on the dependent variables; however, they all had a significant and positive effect on PE, in total explaining 30.1% of the variance in PE. Altogether, the direct and indirect effects in the model explained 8.9% and 13.1% of the variation in CHL-C1 and CHL-C2, respectively. The majority of this came from the direct effect of PE, as can be derived by squaring the standardized regression coefficients (R²_PE→_CHL−C1 = 0.070, R²_PE→_CHL−C2 = 0.123).

Participation in the study was based on willingness to take part and depended on parent's consent. This could lead to a skewed selection from the population. Our sample consisted of an approximately equal number of boys and girls, and an equal number of 8th, 9th and 10th graders. The proportion of sport active youth in our sample were 67%, while estimates in other studies vary. A national survey from 2011 found that 61% of 15-year-olds were active members of sport clubs (40), while more recent publications found that 75% participate at some time during their youth (age 13–18 years old) (62). Another nationally representative survey found that 59% of all lower-secondary school students are active in sport clubs in 2021, however the proportion have ranged from 66% to 59% in the years between 2010 and 2022 (63). This means that our sample might be slightly skewed with a higher proportion of sport active youth, however the differences are small.

The stepwise and transparent analysis applied in the present study is a strength. We utilized scales for CHL that have been developed according to scientific standards (31, 36) and followed best practice recommendations for SEM studies and CFA, including appropriate estimation techniques and fit evaluation (43, 44). We could account for measurement error directly in the estimated models, and the excellent fit of the final model supports the underlying hypothesis. However, the cross-sectional design of the study did not allow us to draw conclusions about the causality of the relationships examined. Possibly it is the higher levels of CHL-C that explain academic achievement in PE, although a reciprocal relationship between success in PE and CHL-C abilities seems likely. Also, whether these abilities are acquired through learning in PE or elsewhere cannot be determined with certainty. There will always be variables not included in the model that could influence complex social constructs, such as CHL. The fact that we did not include measures of the first and second domain of CHL in this study could be considered a limitation. It is possible that students who demonstrate confidence in their interactive and democratic abilities may not necessarily possess the ability to critically appraise health-related information. However, we would argue that fostering these abilities is valuable regardless of their connection to cognitive abilities to appraise health information. In a previous study, we found a moderate association between the first and third domain of CHL (36). Future studies should include objective indicators of CHL, particularly in the domain of critical information appraisal (CHL-A), as demonstrating high confidence in one's ability to judge the credibility of information could be interpreted as an indicator of being non-critical.

Our indicator of academic achievement in PE was the expected grade. In practice, certain behaviors and competencies are rewarded with higher grades. Although grades are based on the same national curricula standards in all Norwegian schools, there are local variations in teachers’ assessment competence (64), interpretations of curricula standards, and, subsequently, what teachers emphasize when grading student achievement in PE. In addition, we could not access the actual grades of students; instead, we asked what grade they expected to get in PE the current semester. Although this is an additional weakness, research has shown that self-reported grades are positively correlated with actual grades and that these are often based on past evaluations, along with an optimistic prediction of future results (65). We also compared mean expected grades for the 10th graders in our sample with the population means for six subjects (Mathematics, Norwegian Language Arts, Social studies, Food and Health, Science and PE) (53). In all subjects the expected grades in our sample were slightly higher than the actual national grades, the differences in means ranged from 0.2 in social studies to 0.4 in mathematics and PE. Differences are likely due to optimistic expectations of grades; however, we cannot rule out that our sample are slightly skewed with a higher proportion of high-achieving students than the population in general.

We utilized only self-reported (subjective) indicators of CHL in the present study. There is generally more measurement error associated with self-measurement as opposed to performance-based (objective) indicators (66). The benefits are, among other things, that they are easy to administer and are experienced as less burdensome for participants to complete (67). Also, an important challenge with objective indicators of CHL is the emphasis this puts on cognitive and functional abilities (e.g., reading and understanding text). This was less relevant in a PE context. With self-reported indicators it is easier to capture different dimensions of complex constructs. In addition, it is well documented that self-efficacy can predict behavior (68). Adding objective indicators could have enhanced the validity of our results, but at the time of our study, performance-based indicators for assessing all aspects of adolescents’ CHL were not yet available (31). Recently, a performance-based instrument for HL encompassing all three domains of CHL has been published (69). The instrument showed promising psychometric properties and could be applied along with self-reported indicators developed in the Norwegian school context (e.g., 36) in upcoming studies. Subsequent investigations should also examine how teaching health in PE and other subjects can contribute to facilitating learning in all aspects of CHL, including the appraisal of health-related information and understanding the social structures that impact health opportunities for all.

We have examined the relationship between academic achievement in PE and the third domain of CHL. The results indicated a positive relationship, and the present practice of PE in our sample seems to nurture CHL-C abilities. The present study contributes to the ongoing discussion on the role of health within the school subject of PE and, subsequently, how PE contributes to HLS and health education in general. We have argued that health is an object of learning in PE and have used the concept of CHL to illuminate and empirically investigate some resources for health that are developed within PE. Importantly, CHL calls attention to the collective and individual attributes for health, which can help counteract an individualistic focus in school-based health education within PE and across different subjects. As such, CHL complements a resource-based health perspective, underpinning the need for student-active approaches that enable content negotiation, action, and reflection regarding matters concerning health and well-being.