Indoor air problems and the perceived social climate in schools : A multilevel structural equation analysis

1 1 2 3 4 This is the post print version of the article, which has been published in Science of the Total Environment 2018, 624, 1504–1512. The final publication is available at https://doi.org/10.1016/j.scitotenv.2017.12.126 5 6 7 Running head: IAQ and social climate in schools 8 9 10 Indoor air problems and the perceived social climate in schools: a multilevel structural equation 11 analysis 12 13 Eerika Finell* 14 Asko Tolvanen 15 Ulla Haverinen-Shaughnessy 16 Seppo Laaksonen 17 Sakari Karvonen 18 Reijo Sund 19 Pauliina Luopa 20 Juha Pekkanen 21 Timo Ståhl 22 23 *Requests for reprints should be addressed to Eerika Finell, Faculty of Social Sciences, Linna, 33014 24 University of Tampere, Tampere, Finland. E-mail: [email protected] 25 26 27 2 Other affiliations: 1 2 Asko Tolvanen: Methodology Centre for Human Sciences, P.O. Box 35, 40014 University of 3 Jyväskylä, Finland 4 Ulla Haverinen-Shaughnessy: Department of Health Protection, National Institute for Health and 5 Welfare, P.O. Box 95, 70701 Kuopio, Finland 6 Seppo Laaksonen: Department of Social Research, P.O. Box 68, 00014 University of Helsinki, 7 Finland 8 Sakari Karvonen: Department of Health and Social Care Systems, National Institute for Health and 9 Welfare, P.O. Box 30, 00271 Helsinki, Finland 10 Reijo Sund: 1) Centre for Research Methods, Department of Social Research, P.O. Box 18, 00014 11 University of Helsinki, Finland, 2) Institute of Clinical Medicine, University of Eastern Finland, 12 Kuopio, Finland 13 Pauliina Luopa: Department of Welfare, National Institute for Health and Welfare, P.O. Box 30, 14 00271 Helsinki, Finland 15 Juha Pekkanen: 1) Department of Public Health, P.O. Box 20, 00014 University of Helsinki, Finland, 16 2) Department of Health Protection, National Institute for Health and Welfare, Finland 17 Timo Ståhl: Department of Welfare, National Institute for Health and Welfare, Biokatu 10, 18 33520 Tampere, Finland 19 20 21 3 Abstract 1 Indoor air problems in schools appear to influence learning outcomes and absence rates. 2 However, previous research has not investigated whether indoor air problems influence the social 3 climate of schools. Therefore, we studied whether indoor air problems observed in schools 4 associate with students’ perceptions of the teacher-student relationship and class spirit. The 5 nationwide sample of Finnish schools (N = 194 schools/27 153 students) was analyzed using 6 multilevel structural equation modeling. Data on the schools’ social climate collected from 7 students were merged with independently collected data on observed indoor air problems from 8 school principals. We found that the teacher-student relationship was reported to be worse in 9 schools with observed indoor air problems compared to those without observed indoor air 10 problems. Furthermore, the reported class spirit was worse in schools with observed indoor air 11 problems, but only among students with a high grade point average. Our findings indicate that 12 indoor air problems may affect the student-perceived social climate. 13 Keywords (6): indoor air quality, social climate, teacher-student relationship, psychosocial effect, 14 lower secondary school, class spirit 15 16 17 18 19 20 21 22 23 4 1. Introduction 1 Problems in schools’ indoor air quality (IAQ) are widespread. For example, some 24% of schools in 2 Finland, 20% in the Netherlands, and 40% Spain were estimated to have moisture problems, such 3 as dampness or mold. 1 Furthermore, many studies have found that a large proportion of schools 4 are inadequately ventilated. 2–4 Such problems can increase the risk to students’ health, 5,6 5 absence rates, 7,8 and decreased academic performance (e.g., standardized test scores). 3,7,9 6 Despite these findings, little is known about whether and how poor IAQ associates with the 7 students’ perception of social climate in schools. The social climate reflects, for instance, the goals, 8 norms, interpersonal relationships and teaching practices, organizational patterns, and school 9 facilities present in a specific school. 10,11 Poor student-perceived social climate associates with 10 students’ decreased well-being and academic performance, 10–13 lower self-esteem 14 and 11 problems in their school adjustment 15,16, for example. Given that the perceived social climate 12 plays an important role in students’ psychosocial well-being and learning, it is essential to study 13 how schools’ IAQ problems relate to it. The need to study the effects of IAQ on the social climate 14 in schools is acknowledged in the “indoor air research” literature as well. 17 15 16 1.1. Student-perceived social climate and indoor air problems in schools 17 Although there is no previous research assessing whether and how schools’ IAQ problems relate to 18 the student-perceived social climate, some indirect evidence suggests that there may be an 19 association between them. Previous research has associated schools’ IAQ problems with increased 20 rates of teachers’ sick leave and decreased well-being, 18, 19 and it may induce interpersonal 21 conflicts as well as anxiety and fear.20 Furthermore, it has been shown that teachers, who perceive 22 their school’s physical environment unsatisfactory, report more negative attitudes and lower 23 5 moral toward their work and notice more problems in the school’s social climate than teachers 1 who perceive the environment as satisfactory. 21–23 Finally, also principals’ perceptions of their 2 school’s physical environment has been associated with their understanding of the school’s ability 3 to deliver instruction.24 4 However, it is unlikely that IAQ problems influence only school personnel. First, given that 5 school’s organizational culture and teachers’ well-being and attitudes relate to the student6 perceived social climate, 25–28 it is plausible that also students in schools with IAQ problems report 7 lower social climate than students in schools without such problems. Furthermore, previous 8 research indicates that also students’ perceptions of their school’s physical environment associate 9 with how they perceive the school’s social climate. 11,29 Finally, students’ absence rates have been 10 found to be higher in schools with poor IAQ than in schools without such problems, 8 which may 11 also reflect issues in the social climate, since the social climate associates with school 12 absenteeism. 30 Based on the evidence, it appears that schools’ IAQ problems may relate to factors 13 which constitute the student-perceived social climate, such as perceived social support, discipline, 14 and peer relationships, in many ways. 15 16 1.2. The present study 17 In this study, we focus on two highly relevant factors of the student-perceived social climate: 18 perceived quality of teacher-student relationship and perceived class spirit (i.e., the classroom’s 19 emotional climate). 10,31 These factors are very influential for students’ academic performance, 20 school engagement, health, and health-related behaviors, for example. 32–34 Therefore, it is 21 important to know whether and how IAQ problems in schools relate to these factors. Our first 22 research question asks whether the perceived teacher-student relationship and class spirit among 23 6 lower secondary school students differ between schools with observed IAQ problems and schools 1 without such problems. Based on the literature cited above, 18,20,22,26 we hypothesize that students 2 in schools with IAQ problems report worse perceived quality of the teacher-student relationship 3 and class spirit than students in schools without such problems (H1). 4 Our second research question asks whether the associations between student-level 5 characteristics (e.g., gender) and the student-perceived social climate are different in schools with 6 and without IAQ problems. The need to study this issue is based on previous research showing 7 that students’ perceptions of their school’s social climate is associated with many student-level 8 characteristics, such as gender, socioeconomic status, ethnicity, social support, and academic 9 performance; 13,35,36 different students perceive the social climate of their school differently. Given 10 that not only student characteristics but also contextual factors can have effects on the student11 perceived social climate, 37 we also tested whether the association between school-level 12 characteristics (e.g., school size) and the student-perceived social climate are different in these 13 two school contexts. Analyzing these interactions was considered important since is provides 14 information on whether there are certain types of students or schools that are especially 15 susceptible to report poor teacher-student relationship or poor class spirit when facing IAQ 16 problems. Given that there are no previous studies considering this issue, no hypothesis is posed. 17 18 19 7 2. Material and methods 1 2.1. Data and participants 2 The data were obtained from two sources: a) the School Health Promotion Study (SHP) 2013, 3 focusing on comprehensive school students in grades eight and nine (14–16 years old) and b) the 4 Benchmarking System of Health Promotion Capacity Building (BSHPCB) data collections from 5 comprehensive schools in 2013 (for more details, see Finell et al., 2017). 38 The questionnaire of 6 BSHPCB is usually filled in by the school’s principals. Two variables from BSHPCB were used in the 7 analysis. The first variable measured if and when the triennial inspection required by Health Care 8 Act 1326/2010 had been carried out in the school. This regulatory inspection is done in co9 operation with the representatives of the school (e.g., a principal), the health authorities, and 10 authorities responsible for construction and maintenance of the school building. It includes a 11 number of factors (not just IAQ related) which can influence the well-being of the school 12 community. The inspection of the indoor environment includes review of existing documents, such 13 as results from possible questionnaires on IAQ and health, IAQ measurements, and other building 14 related reports as well as a building walkthrough focusing on areas where problems have been 15 found or suspected (see Finell et al., 2017; Hietanen-Peltola and Korpilahti, 2015). 38, 39 16 The second variable measured whether or not there were biological exposures observed in 17 the school during this regulatory inspection (see 2.2.2.). BSHPCB questionnaire also asks about 18 physical and chemical exposures. However, we focus only on the item measuring biological 19 exposures here because these issues, especially dampness and mould (“mouldy schools”) have been under 20 intensive public concern since 1990’s in Finland. In the future, it would be interesting to perform comparative 21 analyses on the other two items to test whether the associations with schools’ social climate would be 22 different. For the analysis, we included only those schools where a) the inspection was carried out in 23 8 2012 or 2013 and b) where 1) biological exposures were identified during the inspection but the 1 problems had not been remediated or 2) no indoor air problems were identified. This means that 2 73 schools (N=10181 students, 27%) were excluded from the group of schools where the regulatory 3 triennial inspection was carried out in 2012 and 2013. In these schools a) the biological exposures 4 were not included in the inspection, b) problems were found and they were already corrected or c) 5 no information were provided (see Measures). Sixteen schools with fewer than 10 students were 6 also excluded from the analysis (total of 66 students, 0.2%). In addition, respondents who reported 7 their age as at least two years younger or three years older than the average age of their classmates 8 were excluded from the analysis (n= 64, 0.2%). Two schools with 16 and 12 students were excluded, 9 because they were considered to be school-level outliers (in the teacher-student relationship). The 10 final data consist of 27153 students from 194 schools. About 63 per cent of the students were from 11 schools without observed indoor air problems (17127 students from 127 schools). 12 13 2.2. Measures 14 2.2.1. Outcome variables 15 The perceived quality of the teacher-student relationship was measured by three items: “Teachers 16 are interested in how I am doing”, “Teachers treat us, the students, fairly”, and “The opinions of 17 students are taken into consideration in the development of school work”. The response scale was 18 1 = fully agree, 2 = agree, 3 = disagree, and 4 = fully disagree. These items have been used in many 19 previous studies as indicators of the teacher-student relationship. 27,40,41 We modeled teacher20 student relationship as a latent factor with these three items. The standardized loadings of the 21 confirmatory factor analysis are presented in Table 2. The fit indicators are reported in Result 22 section (see also Supporting information). The Cronbach alpha was 0.63. 23 9 The class spirit was measured by three items: “The students in my class get along well”, “The 1 classroom discipline in my class is good” and “The mood in our class is such that I dare to express 2 my opinion freely”. The response scale was the same as above. These items have been used 3 previously by Karvonen and colleagues 40 as an indicator of class spirit. As above, we modeled class 4 spirit as a latent factor. The standardized loadings are presented in Table 2. The fit indicators are 5 reported in Result section (see also Supporting information). The Cronbach alpha was 0.67. The 6 data source of the outcome variables was the SHP. 7 8 2.2.2. Predictor 9 The BSHPCB survey item “Were the following issues evaluated in the most recent inspection: 10 Exposure to biological agents (indoor air, mold, etc.)” had the following response options: no data 11 available; not included in the inspection; inspected, no deficiencies detected; inspected, 12 deficiencies detected but not yet corrected; inspected, deficiencies detected and corrected. In this 13 study, we focus only on the following options: 0 = inspected, no deficiencies detected and 1 = 14 inspected, deficiencies detected but not yet corrected. The resulting variable is referred to as 15 “observed IAQ problems”. 16 17 2.2.3. Background variables 18 Gender (0 = boy, 1 = girl), age, perceived health, parental support, and education of the mother 19 were included only at the student-level (see Statistical analyses and Figures 1 and 2 below). 20 Perceived health was measured by the item: “How is your health in general?” The response scale 21 was 1 = very good, 2 = fairly good, 3 = moderate, and 4 = fairly or very poor The scale was 22 categorized into two categories (0 = very good, 1 = not very good), because only a small 23 proportion reported moderate or bad health. The similar kind of categorization has been used in 24 16 3.1. The association between observed IAQ problems and the student-perceived schools’ social 1 climate 2 First, we tested the fit of each measurement model (see Figures 1 and 2) by carrying out a two3 level confirmatory factor analysis for each variable separately. In order to do that we fixed the 4 loadings equally at the student and school-levels. 47 The model fits were good both for the 5 teacher-student relationship (RMSEA = 0.01, CFI = 1.00, SRMR: student-level = 0.00, SRMR: school6 level = 0.08) and class spirit (RMSEA = 0.01, CFI = 1.00, SRMR: student-level = 0.00, SRMR: school7 level = 0.05). The standardized factor loadings are presented in Table 2 (see further information on 8 measurement models and the fit indicators in Supporting information). 9 Table 2. Standardized loadings of the teacher-student relationship and class spirit latent 10 variables (student-level: N = 27018–27025/ School-level: N = 194). 11 Teacher-student relationship Class spirit Student-level Student-level λ λ The opinions of students are taken into consideration in the development of school work. 0.65* The students in my class get along well. 0.73* Teachers are interested in how I am doing. 0.56* The classroom discipline in my class is good. 0.54* Teachers treat us, the students, fairly. 0.60* The mood in our class is such that I dare to express my opinion freely. 0.63* School-level School-level The opinions of students are taken into consideration in the development of school work. 0.91* The students in my class get along well. 0.87* Teachers are interested in how I am doing. 0.67* The classroom discipline in my class is good. 0.57* Teachers treat us, the students, fairly. 0.71* The mood in our class is such that I dare to express my opinion freely.a 1.00* * p < 0.001 12 a The residual variance is set to zero at the school-level. 13 14 17 There was a statistically significant variability both in the latent teacher-student relationship 1 - variable within (σ2W = 0.204, p < 0.001) and between schools (σ2B = 0.010, p < 0.001) and in the 2 latent school spirit - variable within (σ2W = 0.276, p < 0.001) and between schools (σ2B = 0.012, p < 3 0.001). The ICC of the both latent variables were 0.05 and 0.04 respectively, meaning that about 4 4–5% of the variability occurred between schools. Although the ICCs were relatively small, they 5 had strong design effects (DEFF;61 teacher-student relationship: DEFF = 7.46; school spirit: DEFF = 6 6.76). 7 Then we estimated two random intercept models (i.e., Model 1a and Model 1b) in order to 8 test whether the perceived teacher-student relationship and class spirit differed between schools 9 with and without observed IAQ problems (see Table 3). The model fits were good both for the 10 teacher-student relationship (Model 1a: RMSEA = 0.02, CFI = 0.97, SRMR: student-level = 0.01, 11 SRMR: school-level = 0.07, see Table 2) and for school spirit (Model 1b: RMSEA = 0.02, CFI = 0.99, 12 SRMR: student-level = 0.01, SRMR: school-level = 0.06, see Table 2). After all background variables 13 were inserted into Model 1a, observed IAQ problems associated significantly with the teacher14 student relationship. The effect size was 0.5, indicating a medium-sized effect 57. However, after 15 all background variables were inserted into Model 1b, observed IAQ problems did not associate 16 significantly with school spirit (see further information on univariate associations in Supporting 17 information) . 18 In Model 1a, the most important student-level background variables were parental support, 19 perceived health, and GPA. The most important school-level background variables were ethnic 20 composition (measured by students’ self-reported ethnic background) and the school’s average 21 GPA. In Model 1b, the most important student-level background variables were parental support, 22 gender, and perceived health. The most important school-level background variables were school 23 18 size and the school’s average GPA (see Table 2). Model 1a explained 11% of the total variance at 1 the student-level (R2 = 0.114) and 22% at the school-level (R2 = 0.222). Model 1b explained 8% of 2 the total variance at the student-level (R2 = 0.077) and 16% at the school-level (R2 = 0.157), 3 respectively (see Table 2). 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 19 Table 3. Parameter estimates for the teacher-student relationship and the classroom’s social 1 climate1 as a function of studentand school-level variables: A random intercept model 2 (N=27153 students/194 schools). 3 Model 1a Teacher-student relationship Model 1b Class spirit B (SE)a β (SE)b B (SE)a β (SE)b Student-level Genderc -0.06 (0.01)*** -0.07 (0.01)*** 0.13 (0.01)*** 0.13 (0.01)*** Aged 0.01 (0.01) 0.02 (0.01) -0.02 (0.01) -0.02 (0.01) Age squarede -0.06 (0.01)*** -0.06 (0.01)*** Perceived healthf 0.08 (0.01)*** 0.09 (0.01)*** 0.16 (0.01) *** 0.15 (0.01)*** Parental supportg 0.12 (0.00)*** 0.26 (0.01)*** 0.07 (0.01)*** 0.13 (0.01)*** Parental support squarede 0.03 (0.00)*** 0.06 (0.01)*** 0.02 (0.01)*** 0.03 (0.01)*** Education of mother h 0.05 (0.01)*** 0.06 (0.01)*** -0.02 (0.01) -0.01 (0.01) Grade point averagei -0.06 (0.01)*** -0.11 (0.01)*** 0.01 (0.01)* 0.02 (0.01)* Grade point average squarede -0.02 (0.01)* -0.02 (0.01)* Ethnic backgroundj -0.02 (0.03) -0.01 (0.01) 0.08 (0.03)** 0.03 (0.01)** School-level Observed IAQ problemsk 0.05 (0.02)** 0.25 ( 0.08)** 0.02 (0.02) 0.08 (0.08) Grade point averaged -0.22 (0.09)*l -0.28 (0.10)** -0.24 ( 0.09) ** -0.27 (0.10)** Ethnic background -0.36 (0.12)** -0.19 (0.09)*n 0.03 (0.19) 0.02 (0.09) School size o 0.01 (0.00)m 0.16 (0.08)* 0.01 (0.01)* 0.24 (0.09)* σ2W 0.17 (0.01) 0.89 (0.01) 0.24 (0.01) 0.92 (0.01) σ2B 0.01 (0.00) 0.78(0.07) 0.01 (0.00) 0.84 (0.08) *p < 0.05, **p < 0.01, *** p < 0.001 4 1 A higher value in the teacher-student relationship and the classroom’s social climate means more problems. 5 a Unstandardized beta 6 b Standardized beta 7 c Scale 0–1 (0 = boy) 8 d Grand mean centered 9 e Quadratic term. Only if the quadratic term was significant (p < 0.05), it was included in the final model and reported. 10 There were no significant quadratic terms at the school-level. 11 f Scale 0–1 (0 = very good health) 12 20 g Scale 1–4 (1 = good support, grand mean centered) 1 h Scale 0–1 (1 = university education) 2 i Scale 0–2 (0 = GPA less than 7.5, grand mean centered) 3 j Scale 0–1 (0 = one or both parents are born in Finland) 4 k Scale 0–1 (0 = no observed IAQ problems in the school) 5 l p = 0.012 6 m p = 0.064 7 n p = 0.027 8 o In these models the scale was changed as follows: school size/100. 9 10 3.2. The effect of the school’s IAQ problems on the associations between the background and 11 outcome variables 12 We tested differences on the associations between a) student-level characteristics and the 13 student-perceived social climate and b) school-level characteristics and the student-perceived 14 social climate in schools with and without IAQ problems. In order to test these differences we 15 used multigroup two-level structural equation modeling and Wald-test. The association between 16 student-level GPA and class spirit was moderated by observed IAQ problems (Wald-test = 6.04, p = 17 0.014). This association was stronger in schools with IAQ problems (standardized beta = 0.05, p < 18 0.001) than in schools where there were no such problems (standardized beta = 0.00, p = 0.821; 19 see further information on univariate associations in Supporting information). In order to better 20 understand this result, we ran three separate analyses. We tested students whose GPA was a) less 21 than 7.5, b) between 7.5 and 8.4 and c) over 8.4. Except for GPA, the other background variables 22 in these three analyses were the same as in Model 1b (see Table 2). Students with the highest GPA 23 (i.e., over 8.4) perceived more problems in class spirit in schools with observed IAQ problems than 24 in schools without such problems (standardized beta = 0.19, p = 0.032). The effect size is 0.4, 25 indicating a small-sized effect. 57 There were no differences between schools among students in 26 the lower GPA clusters (GPA 7.5–8.4: standardized beta = 0.12, p = 0.193; GPA < 7.5: standardized 27 beta = 0.01, p = 0.948). 28 21 1 Figure 3. The mean difference of class spirit between schools with and without observed IAQ 2 problems divided by the standard deviation of class spirit. 3 0 = GPA < 7.5, p = 0.948; 1 = GPA 7.5–8.4, p = 0.193; 2= GPA > 8.4, p = 0.032 (N = 193–194 4 schools). 5 6 Furthermore, the linear terms of the parental support and teacher-student associations were 7 significantly different (Wald-test = 4.75, p = 0.029) between the two school contexts (no observed 8 IAQ problems: standardized beta = 0.27, p < 0.001; observed IAQ problems: standardized beta = 9 0.23, p < 0.001; see further information on univariate associations in Supporting information). The 10 Wald-test between quadratic terms was not significant (Wald-test = 0.50, p = 0.479). This small 11 effect indicates that the difference in the perceived quality of the teacher-student relationship 12 between the two school contexts was larger among students with high parental support than 13 students with low parental support. In total, students with low parental support, however, 14 evaluated their teacher-student relationship more negatively than students with high parental 15 support. 16 17 18 Grade point average Class spirit 22 4. Discussion 1 Our main finding was that on average, students in schools with observed IAQ problems reported 2 worse teacher-student relationship than students in schools without such problems. Furthermore, 3 we found that students with a high GPA reported worse class spirit in schools with IAQ problems 4 than in schools without such problems. 5 Previous research has reported associations between poor IAQ in schools and both health 6 and learning outcomes (e.g., respiratory symptoms, illness absence, and academic performance). 7 3,7,9,62 Our findings contribute to this research by showing that IAQ problems in schools may have 8 an effect on the student-perceived social climate. To our knowledge, this is the first time when 9 these effects have been assessed. Given that social climate is shown to influence various factors, 10 such as academic performance, absence rates, and students’ psychosocial well-being, 10,30,31 our 11 findings are important. Furthermore, we found that a school’s observed IAQ problems significantly 12 modified the associations between student-level characteristics and the student-perceived social 13 climate. Although these associations were weak, the findings supported the previous research by 14 Finell and colleagues38 by concluding that schools with and without IAQ problems may represent 15 two different social contexts. They also suggest that the social and psychological consequences of 16 IAQ problems in schools result from complex processes where both the students’ and school 17 personnel’s characteristics play a role. 18 An important task for future research is to examine why IAQ problems in schools may 19 influence the student-perceived social climate. Our results and the literature provide three 20 potential intertwined paths. The first path relates to the impact of poor IAQ on school personnel, 21 for example by decreasing their well-being, increasing sick leave, and causing possible problems in 22 management and decision-making procedures. 19,20,26,28,63 The second path relates to the impact of 23 23 poor IAQ on students. The students’ perceived health is associated with the student-perceived 1 social climate, 10 as was also found in our study. The third path relates to the poor condition of the 2 school facilities as well as anticipated remediation of the school building. 22,23 In Finland, school 3 general management and facility management are independent from each other. In other words, 4 school principals and staff do not have a direct role or oversight of the facility management. They 5 can report any issues or concerns but the allocation of funds for school facility management and 6 remediation decisions are made by the municipality technical board. However, remediation 7 processes often result in disturbances (e.g., moving to temporal buildings and/or noise), which can 8 be stressful both to school personnel and students, and may therefore influence the student9 perceived social climate. 10 From a practical point of view, an important question is how it is possible to prevent the 11 potential social consequences caused by IAQ problems in schools. The main task is naturally to 12 investigate and properly remediate indoor air problems in the school building. However, these 13 processes are often slow. In case the remediation is delayed, the prevention of health problems 14 among teachers and students can involve source control (e.g., more frequent cleaning), enhanced 15 ventilation, and exposure prevention (e.g., not using problematic areas). Furthermore, 16 management and decision-making procedures should be developed in order to avoid conflicts 17 inside the school organization and experiences of injustice.64 Open communication and sharing 18 up-to-date (regular, factual) information with school personnel, students, and parents should help 19 to maintain trust between parties. 20 The strengths of the study include the large and representative sample. The use of multilevel 21 structural equation models has many advantages, including better statistical power compared to 22 models using aggregated school-level data and avoiding spurious findings from using student-level 23 24 data without taking the hierarchical data structure into account. 48 The limitations of the study 1 include the cross-sectional design and the use of questionnaire data. Furthermore, we did not 2 have physical measures. This means that we have to rely on principles report on the findings of 3 triennial inspection on biological exposures and that we are cannot better distinguish schools 4 based on the factors that have caused their IAQ problems. It is likely that there exists variation in 5 the severity of schools’ IAQ problems and their social consequences. We also do not know the 6 degree teachers and students were aware of their school’s IAQ problems. In the future, these 7 issues should be further studied using longitudinal data and physical measures. Future research 8 should also detect factors that can mediate the association between school’s indoor environment 9 and the perceived social climate. Also possible building-related confounders (e.g., poor sanitary 10 conditions or unpleasant school environment) should be considered. 11 The present study is part of a larger project aiming to clarify associations between 12 psychosocial factors and IAQ problems among comprehensive school students in Finland. In a 13 previous study, we found associations between both the physical and psychosocial environment 14 and subjective IAQ in schools. 38 In the present study, psychosocial factors are the outcome 15 variables. Our study underlines the fact that the associations between the perception of IAQ 16 problems and psychosocial factors are highly complex and are likely to influence each another in 17 various ways. 18 19 20 21 22 25 5. Conclusions 1 Our study found that problems in a school’s IAQ associated with the student-perceived social 2 climate. Furthermore, we found that the school’s observed IAQ problems significantly modified 3 the associations between student-level characteristics and the student-perceived social climate. 4 Our study points to a need to study schools’ IAQ problems from a multidisciplinary perspective in 5 order to better understand how IAQ problems influence students’ psychosocial well-being. 6 Furthermore, from practical point of view, when evaluating the urgency of remediation, also IAQ 7 problems’ possible social consequences should be taken into account. Fast and effective 8 remediation process not only prevent adverse health effects but can also prevent problems in 9 school’s social climate. 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 32 59. Rubin DB. Inference and missing data. Biometrika. 1976;63:581–592. 1 60. Enders C, Bandalos D. The relative performance of full information maximum likelihood 2 estimation for missing data in structural equation models. Struct. Equ. Model. Multidiscip. J. 3 2001;8:430–457 . 4 61. Ukoumunne OC, Gulliford, MC, Chinn S, et al. Methods for evaluating area-wide and 5 organisation-based interventions in health and health care: a systematic review. Health 6 Technol. Assess. 1999;3. 7 62. Mendell MJ, Eliseeva EA, Davies MM, Lobscheid A. Do classroom ventilation rates in California 8 elementary schools influence standardized test scores? Results from a prospective study. 9 Indoor Air. 2016;26:546–557. 10 63. Miller RT, Murnane RJ, Willett JB. Do teacher absences impact student achievement? 11 Longitudinal evidence from one urban school district. Educ. Eval. Policy Anal. 2008; 30: 181– 12 200. 13 64. Finell E, Seppälä T. Indoor air problems and experiences of injustice in the workplace: A 14 quantitative and a qualitative study. Indoor Air. 2017; doi:10.1111/ina.12409. 15 16 17 18 19 20 21 22 23 24 25 33 1 2 Figure 1 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 = measurement model 33 34 Figure 1. Hypothesized structural equation model: The teacher-student relationship as a latent 35 outcome variable. 36 Teacher-student relationship Teachers are interested in how I am doing Gender Teachers treat us, the students, fairly The opinions of students are taken into consideration in the development of school work Age Perceived health Parental support GPA Ethnic background Education of mother Teacher-student relationship Student-level School-level Teachers are interested in how I am doing Teachers treat us, the students, fairly The opinions of students are taken into consideration in the development of school work GPA Ethnic background School size Observed IAQ problems 34 1 Figure 2 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 = measurement model 32 33 Figure 2. Hypothesized structural equation model: Class spirit as a latent outcome variable. 34 Class-spirit Student-level School-level The students in my class get along well. The classroom discipline in my class is good. The mood in our class is such that I dare to express my opinion freely. GPA Ethnic background School size Observed IAQ problems Gender Age Perceived health Class-spirit Parental support The students in my class get along well. The classroom discipline in my class is good. The mood in our class is such that I dare to express my opinion freely. GPA Ethnic background Education of mother