Linking Health, Comfort and Indoor Environmental Quality in Classrooms with Mechanical Ventilation or Window Airing: A Controlled Observational Study

Abstract

Effective classroom ventilation is essential for indoor environmental quality (IEQ), comfort and health of schoolchildren, who spend substantial time indoors. This controlled observational study compared manual window airing (WA) with decentralized mechanical ventilation (DV) in six classrooms of two elementary schools during the winter infection period. Symptoms of upper respiratory tract infections, salivary biomarkers, well-being, perceived comfort, and classroom-level IEQ were assessed through questionnaires, saliva samples and long-term monitoring. Ninety-eight schoolchildren participated (64 WA, 34 DV). Symptom-based outcomes of the WURSS-K questionnaire showed consistently lower illness burden in group DV, with several parameters reaching statistical significance and an absolute risk reduction of 7.8%. Salivary immunoglobulin A (sIgA) concentrations were also significantly lower in group DV (approximately 39–59%, p ≤ 0.01). Sensitivity analyses showed positive associations of CO₂ and PM_2.5 with sIgA and indicated that PM_2.5 exposure accounted for group differences. Comfort perceptions mirrored measured IEQ: DV classrooms exhibited warmer, more stable thermal conditions, lower CO₂ and PM_2.5, and slightly better thermal and draught-related impressions. Overall, decentralized mechanical ventilation supported favorable IEQ and comfort and may influence mucosal immune activity through reduced particulate exposure, complementing the observed reduction in symptom burden. A multidimensional approach integrating medical outcomes with continuous IEQ monitoring proved valuable and should be expanded in larger, balanced cohort studies.

1. Introduction

Schoolchildren spend a large part of their day in school buildings, where indoor environmental quality (IEQ)–including indoor air quality (IAQ) and thermal comfort–shapes their exposure to indoor pollutants, airborne pathogens, and thermal conditions. Indoor carbon dioxide (CO₂) concentrations in classrooms primarily increase due to human respiration, as exhaled air contains substantially higher CO₂ levels than outdoor air, which typically exhibits a background concentration of approximately 420 ppm [1]. Consequently, indoor CO₂ serves as a practical indicator of ventilation adequacy and the balance between occupant-related emissions and outdoor air supply. CO₂ concentrations below 1000 ppm are generally considered hygienically safe, whereas concentrations above 2000 ppm are classified as unacceptable [2,3]. The effectiveness of ventilation in improving IAQ depends not only on air exchange rates but also on the quality of the supplied outdoor air. While outdoor air is typically cleaner with respect to CO₂ and airborne infectious particles, its particulate matter (PM) load may vary substantially depending on the surrounding environment. Mechanical ventilation systems can reduce indoor exposure to outdoor particulate matter through air filtration, whereas natural window airing introduces outdoor air without filtration. Reports from the German Environment Agency indicate that pupils and teachers frequently experience headaches, fatigue, lack of concentration, eye irritation, and upper respiratory tract symptoms associated with time spent in school environments [4]. Several international studies show that ventilation rates in school classrooms are often inadequate, as indicated by CO₂ concentrations frequently exceeding 1000 ppm [5,6,7,8]. This problem is particularly pronounced in classrooms without mechanical ventilation during the heating season, when windows tend to remain closed to maintain thermal comfort [9]. A review of 125 studies in naturally ventilated primary schools found that 81% of classrooms exceeded the 1000 ppm CO₂ threshold, with CO₂ correlating with PM, volatile organic compounds (VOCs), and microbial loads [5]. A European research project reported that poor IAQ in schools (including inadequate ventilation, elevated CO₂, VOCs, particulate matter, and indicators of dampness) was associated with respiratory symptoms and reduced well-being in children [7].

Traditionally, ventilation research in schools has emphasized cognitive performance, with lower CO₂ repeatedly linked to improved attention and academic outcomes [10,11], and comfort-related aspects of IEQ, such as temperature, humidity, draught, and acoustics. Comfort studies show that ventilation mode and window-opening behavior shape thermal sensation and perceived draught [12,13]. Beyond performance and comfort aspects, upper respiratory tract infections (URTIs) are highly prevalent worldwide, usually mild and self-limiting, and most commonly manifest as the common cold [14]. Despite their relevance, only a limited number of school-based studies integrate questionnaires on symptoms of URTIs. The WURSS-K (Wisconsin Upper Respiratory Symptom Survey for Kids) provides a child-appropriate instrument for evaluating acute respiratory illness [15]. A study in Poland linked PM exposure to higher frequencies of respiratory symptoms in over 1400 Polish children [16], a study in northern Thailand reported impaired lung function among primary schoolchildren exposed to high ambient PM_2.5 concentrations [17]. Besides air pollutants, airborne pathogens represent an additional major determinant of respiratory morbidity in classrooms, where high occupancy and limited ventilation facilitate both pollutant accumulation and infection transmission. Interest in ventilation increased markedly during the coronavirus disease 2019 (COVID-19) pandemic, with sufficient air dilution and appropriate distribution recognized as key mitigation strategies [18,19,20,21,22,23]. CO₂ concentrations can serve as an indicator of relative airborne infection risk [24], provided no air purifier is operating in the room. A review on airborne infection risks in schools concluded that IAQ is often inadequate and that reliable natural or mechanical ventilation is essential for mitigating airborne transmission [8]. In classrooms where window airing is the only option, CO₂ traffic lights (CO₂ measuring devices with a visual feedback) are recommended to support adequate ventilation behavior [25]. Mechanical ventilation systems with heat recovery can offer controlled and energy-efficient ventilation, increasing outdoor air exchange while maintaining thermal conditions. Decentralized mechanical ventilation units are increasingly considered a practical retrofit solution, supplying continuous outdoor air without requiring ductwork. In 2021, the German Environment Agency recommended that classrooms be progressively equipped with ventilation and air-conditioning systems for reasons of indoor air hygiene and sustainability [26]. Nevertheless, although Germany has approximately 32,000 general schools [27], it was estimated that fewer than 10% were equipped with mechanical ventilation systems in 2021 [28].

The present controlled observational study investigates differences between manual window airing supported by CO₂ traffic lights and decentralized mechanical ventilation in real-world classroom settings during the winter infection period. A multidimensional approach is applied, combining (i) symptom-based self-assessments of respiratory health, (ii) salivary biomarkers of inflammation and mucosal immunity, (iii) subjective well-being and perceived comfort, and (iv) continuous IEQ monitoring (CO₂, PM_2.5, VOC, temperature, relative humidity). A detailed analysis of the long-term monitoring, including indoor and outdoor environmental data, is presented in [29]. This integrative study design enables health-related outcomes to be interpreted within the context of classroom exposure conditions and supports the identification of potential environmental confounders, such as particulate matter. By linking medical outcomes with continuous environmental monitoring, the study adds a health-centered perspective to school ventilation research. It offers new insights into the potential health and comfort implications of decentralized mechanical ventilation and highlights the value of integrating subjective assessments, physiological markers, and continuous environmental measurements in school-based field studies.

2. Materials and Methods

2.1. Study Design

During the peak infection season of fall/winter 2023/2024, a prospective, controlled observational study was conducted from October 2023 to April 2024 in three elementary schools in Middle Franconia, Germany. The study “Schule, Luft und Gesundheit” (SLG, “School, Air and Health”) received ethical approval from the Joint Ethics Committee of the Universities of Applied Sciences of Bavaria (GEHBa) prior to data collection (ID GEHBa-202307-V-119-R, 13 September 2023). Additional approval was granted by the government of Middle Franconia, as well as by the school authorities and principals of the participating schools. This study was a non-interventional, single-center observational cohort study and therefore did not meet the ICMJE definition of a clinical trial requiring prospective registration. For reasons of transparency and open reporting, the study was retrospectively registered with ISRCTN (ISRCTN14284039) after completion of data collection [30]. Reporting followed the STROBE recommendations for observational studies [31].

The study consisted of two parts: an assessment of students’ health and comfort, and a long-term monitoring of IEQ parameters in the classrooms and environmental parameters at each school site. The overall study design follows the methodology described in [29], which reports the results of the long-term monitoring. The current study focuses specifically on the assessment of health and comfort using questionnaires and saliva samples. Given the field-based nature of the study and the integration of repeated health assessments with continuous IEQ monitoring, the achievable sample size was primarily constrained by organizational and participation-related factors rather than determined by an a priori statistical power calculation. Originally, three ventilation concepts were investigated: manual window airing, decentralized mechanical ventilation, and central mechanical ventilation. However, due to the low participation rate in the central ventilation group, this group was not considered in the analysis of students’ health and comfort. In addition, the long-term monitoring revealed that the two classrooms of this group either received central classroom ventilation or a (mostly insufficient) manual window ventilation due to an unknown energy-efficient control strategy of the central ventilation system (see [29]). Therefore, the analysis of health and comfort focuses on the following two groups:

• Window Airing (WA):
  Four school classes in classrooms with manual window airing only and additional CO₂ traffic lights
• Decentralized (mechanical) Ventilation (DV):
  Two school classes in classrooms with one decentralized AHU each (window airing possible)

Window opening was performed by the teachers or pupils according to their usual classroom routines. No standardized or study-defined ventilation protocol was provided, and the study team did not influence or instruct the window-opening behavior. Thus, window opening followed regular school practice and reflects real-world conditions. It should be noted that CO₂ traffic lights were installed only in the window-aired classrooms. The presence of these devices may have influenced window-opening decisions and could therefore introduce behavioral differences between ventilation groups.

2.2. Ventilation Concepts and Classrooms

Group WA includes four classrooms that can be ventilated by manually opening three large south-facing sliding windows (Figure 1a). Each classroom is equipped with a CO₂ traffic light (measuring range: 400–5000 ppm) providing visual feedback according to CO₂ levels (green < 1000 ppm, yellow 1000–2000 ppm, red ≥ 2000 ppm) [32].

In group DV, each of the two classrooms is supplied with fresh outside air through a decentralized air handling unit (AHU) installed at the rear wall (Figure 1b). The supply air enters the room via a ceiling-mounted pipe, with four air outlets. Each AHU provides a maximum airflow rate of 1000 m³/h and is equipped with a counterflow plate heat exchanger for heat recovery (heat transfer only, no humidity recovery). Supply and extract air are filtered with ePM₁ 55% (F7) and ePM₁₀ 60% (M5) filters, respectively [33]. In addition to mechanical ventilation, manual airing via east-facing windows or, in one classroom, a balcony door is possible as well. The AHUs operate according to a time schedule set by the school’s facility manager (Monday to Friday, 7:45 a.m. to 1 p.m.). During day mode, the AHU maintains a constant fan speed and the setpoint for the supply air temperature is 20 °C. From February 2024 onward, the AHU in one classroom is equipped with a CO₂ sensor in the extract air enabling CO₂-based volume flow control during operating hours. Outside the operating hours, the AHUs switch to an energy-saving backup mode. Comprehensive details of all ventilation concepts and classroom designs are provided in [29].

Table 1. Classroom characteristics (WA: Window Airing, DV: Decentralized Ventilation) (table adapted from [29]).

	Classroom
Properties	WA-1	WA-2	WA-3	WA-4	DV-1	DV-2
Length (m)	10.7				8.50
Width (m)	6.7				7.0
Height (m)	3.2				3.0	3.0–4.65
Floor area (m2)	71.7				59.5
Volume (m3)	225.8				178.5	228.8
Number of openable windows	3				6	5/1
Window type	sliding windows				tilt-turn	tilt-turn/door
Orientation	south (180°)				east (90°)
Solar shading	fixed structural				operable external/ fixed structural
Heating system	wall-mounted radiators				wall-mounted radiators
Year of construction	1974				1994
Occupants	23	20	25	23	22	26
Specific area (m2/person)	3.1	3.6	2.9	3.1	2.7	2.3
Specific volume (m3/person)	9.8	11.3	9.0	9.8	8.1	8.8
qtot (perceived air quality) (m3/h)	760	685	811	760	704	805

gives an overview of structural characteristics, occupancy, and required ventilation rates of the six classrooms. Each classroom is labeled according to its ventilation concept (WA or DV) and a consecutive number (1–4 or 1–2) The four classrooms in group WA share identical structural properties, while the two classrooms in group DV differ in height. Room DV-1 has a typical classroom height of 3 m, whereas DV-2 features a gable roof rising from 3 m to 4.65 m, adding 50 m³ of volume. In addition to casement windows, DV-2 also includes an openable balcony door. Group WA represents the oldest school building with the lowest insulation level in this study.

The number of occupants includes the pupils and one teacher, with class sizes ranging from 19 to 25 pupils. All classrooms meet the Bavarian school building regulations requiring at least 2 m² floor area and 6 m³ air space per pupil [34]. The specific area per person ranges from 2.3 m²/person (DV-2) to 3.6 m²/person (WA-2) and the specific volume from 8.1 m³/person (DV-1) to 11.3 m³/person (WA-2). Due to its low occupancy, classroom WA-2 offers the highest space per person in both floor area and room volume. The required total outdoor air ventilation rate q_tot is calculated on the perceived air quality according to DIN EN 16798-1 for a room of category IEQ_II [35]. Owing to varying class sizes, the ventilation rates already differ within the two ventilation groups. Classrooms DV-2 and WA-3 (highest number of persons) require the largest fresh air volume flows of 805 to 811 m³/h.

2.3. Enrollment of Study Participants

Participants were enrolled in September 2023 from the original eight classes included in the study. Schoolchildren were assigned to either one of the two mechanically ventilated groups or the control group based on their existing classroom membership. As classroom allocation was determined by the schools’ internal administrative procedures and not influenced by the study team, group assignment was independent of the children’s health status or individual characteristics, although not randomized in a clinical sense. All participants were in grades 2–4, with a minimum grade level of 2 required to ensure sufficient comprehension for completing the questionnaires.

Participation in this study was voluntary. Inclusion criteria comprised membership of the selected classes, age between 6 and 12 years, and provision of parental consent. Parents were informed during school information sessions prior to the study, and written informed consent was obtained from legal guardians prior to participation.

2.4. Study Procedure and Data Collection

The study was conducted over a seven-month period, from 1 October 2023 to 30 April 2024 (213 days, including 120 school days). Figure 2 provides an overview of the study procedure. In September 2023, information sessions were held with parents to explain the study objectives and procedures, followed by recruiting of study participants. At the beginning of the study, baseline information about the participating children was obtained from parents using a questionnaire. Data collection in schools took place at three time points: mid-November 2023 (T1), late January 2024 (T2), and late April 2024 (T3), with intervals of approximately 10–13 weeks between assessments.

The cold symptom survey WURSS-K was defined as the primary outcome, representing the children’s respiratory symptom burden and serving as the main health-related endpoint of the study. It was completed at home throughout the entire study period. Secondary outcomes comprised (i) salivary biomarker concentrations (sIgA, sCRP), (ii) psychological well-being assessed by the WHO-5 index, and (iii) subjective comfort ratings obtained from the Comfort-13 questionnaire specifically developed for this study. At each time point (T1–T3), these questionnaires and saliva samples were collected during regular school hours under supervision by the study team and in coordination with teachers to minimize disruption of lessons. In addition, long-term monitoring of indoor and outdoor environmental parameters was conducted in all classrooms and at each school site.

2.5. Measurement of Primary Outcome by WURSS-K

Health-related symptoms were assessed using a German version of the “Wisconsin Upper Respiratory Symptom Survey for Kids” (WURSS-K) [36], which evaluates the presence and severity of common cold symptoms and their functional impact. Participating children were instructed to complete the questionnaire on each day they felt ill or experienced cold symptoms, including weekends and vacation days, as well as on the first day they felt recovered and symptom-free. Parents were asked to support their children in completing the questionnaire at home. However, the analysis showed that only a few children completed the questionnaire for the final day of an illness episode when they felt recovered. Therefore, this criterion was excluded from the final evaluation.

The WURSS-K comprises 15 items covering three domains: global illness severity (item 1), symptom severity (six items on common cold symptoms), and functional impact of illness on quality of life (seven items addressing daily activities). Fourteen items are rated on a 4-point ordinal Likert scale (0–3), ranging from 0 (absent or no impairment) to 3 (severe), and forming the global total score (range 0–42). Symptom severity (range 0–18) and functional impact (range 0–21) are calculated as subscale sums. Item 15 assesses perceived change compared with the previous day and is not included in the total score. Scoring follows the procedure described in [15].

As the questionnaire was not previously available in German, a German version of WURSS-K was developed for this study. It was translated according to a standardized translation protocol in consultation with its original developer, Bruce Barrett, and the Wisconsin Alumni Research Foundation (WARF). The use of the finalized German version was approved by WARF. The German translation of the WURSS-K applied in this study is provided in the Supplementary Materials.

2.6. Measurement of Secondary Outcomes

2.6.1. Salivary C-Reactive Protein and Salivary Immunoglobulin A

Saliva samples were collected at each of the three school visits (T1–T3) to assess inflammatory and immunological surrogate parameters. Sampling was performed during the morning hours. Participants were instructed not to eat or drink (except water) for at least 30 min prior to sample collection. Salivary C-reactive protein (sCRP) was quantified using the Salimetrics^® Salivary C-Reactive Protein ELISA Kit (Salimetrics, LLC, Carlsbad, CA, USA), Generation 2 (Item No. 1-2102; Single 96-well kit). Samples were diluted 1:2 in the kit sample diluent. Salivary immunoglobulin A (sIgA) was measured using the DRG Instruments Salivary IgA ELISA (REF: SLV-4636; DRG Instruments GmbH, Marburg, Germany). Samples were diluted in the kit sample diluent according to the instructions. Absorbance was read at 450 nm with a 620–650 nm reference on a standard microplate reader.

2.6.2. WHO-5 Well-Being Index

Psychological well-being was assessed using the WHO-5 Well-Being Index at each of the three sample points (T1–T3). The questionnaire consists of five positively worded items evaluating subjective psychological well-being during the past two weeks [37]. Each item is rated on a 6-point ordinal Likert scale ranging from at no time (0 points) to all of the time (5 points). For evaluation, the raw score is calculated as the sum of the five items and multiplied by four, resulting in a total score ranging from 0 to 100. A score of 0 represents the worst imaginable well-being, whereas a score of 100 indicates the best possible well-being. A value of 50 or below suggests reduced well-being and is often considered an indicator of possible risk for depression [38].

2.6.3. Comfort-13 Questionnaire

At each of the three time points (T1–T3), thermal and acoustic comfort as well as subjective perceptions of well-being were assessed using the Comfort-13 questionnaire, which was specifically developed for this study. The full questionnaire is provided in the Supplementary Materials. It comprises 13 items grouped into five thematic sections. Except for the section on thermal sensation, all scales were designed such that lower scores indicate a more favorable condition (e.g., very alert/very good/no/quiet), with 0 representing the optimal state. Higher scores correspond to decreasing comfort levels. For three of the sections, the textual response options were supplemented by matching smiley icons, which were intentionally presented in color to enhance visual appeal for children.

The first item assesses how awake the child feels, rated on a 5-point ordinal Likert scale ranging from very alert (0) to very tired (4). The second section includes three items evaluating sleep quality, mood, and perceived air quality in the classroom, also rated on a 5-point ordinal Likert scale from very good (0) to very poor (4). The third section addresses draught sensation and perceived humidity, comprising three items rated on a 4-point Likert scale from no (0) to very strong (3). The largest section focuses on thermal sensation and consists of four items assessing the general thermal sensation and potential vertical temperature differences. The response scale is adapted from the seven-point thermal comfort scale defined in DIN EN ISO 7730:2006-05 [39]. A thermally neutral perception is rated as good (0). Cooler sensations are rated with negative values from cool (−1) to too cold (−3), while warmer sensations are rated with positive values from warm (1) to too hot (3). The final section captures acoustic comfort. Here, perceived loudness is rated on a 3-point Likert scale from quiet (0) to very loud (2), and the presence of disturbing noises is recorded as no (0) or yes, the following (1), with an open-ended field for specifying the type of noise.

2.7. Long-Term Monitoring Concept

In each classroom, IEQ parameters were continuously measured using a compact, surface-mounted multi-sensor (FS1600; FuehlerSysteme eNET International GmbH, Nuremberg, Germany) installed at a height of 1.1 m: indoor air temperature, relative humidity (r.h.), CO₂ concentration, VOCs and particulate matter in the fraction PM_2.5. The present position of each window (closed, tilted or open window state) was detected by window contact switches (FTKE-rw and FFG7B; ELTAKO GmbH, Fellbach, Germany; and NodOn SDO-2-1-05; NodOn SAS, Saint-Cry-en-Val, France). The electrical power consumption of the two decentralized AHUs in classrooms DV-1 and DV-2 was measured by current transformers (AT 10 B5; LEM International SA, Meyrin, Switzerland). Meteorological parameters–including ambient air temperature, relative humidity, air pressure, global solar radiation, wind speed and wind direction–were recorded by a weather station (u[sonic]WS7; LAMBRECHT Meteo GmbH, Göttingen, Germany) installed on the roof of each school. In addition, ambient PM_2.5 concentrations were monitored by a PM measuring device (FS1308; FuehlerSysteme eNET International GmbH, Nuremberg, Germany) on the façade of each school building.

Table 2. Installed sensors, measured parameters, measurement ranges and stated accuracies (table adapted from [29]).

Sensor Type	Location	Measurement Parameter	Measurement Range	Accuracy	Source
Multi-sensor device	Indoor	Air temperature	−20 … +50 °C	±0.5 K	[40]
		Relative humidity	0 … 100% r.h.	±3% r.h.
		CO2 concentration	0 … 10,000 ppm	±50 ppm
		PM2.5	0 … 1000 µg/m3	±5 µg/m3
		VOC	0…100%	<10%
Current transformer	Indoor	Electrical current	0 … 10 A	<0.15 A	[41]
Weather station	Outdoor	Ambient air temperature	−40 … +70 °C	±0.1 K	[42]
		Relative humidity	0 … 100% r.h.	±1.5% r.h.
		Ambient air pressure	300 … 1100 hPa	±0.5 hPa
		Solar irradiance	0 … 2000 W/m2	<20 W/m2
		Wind speed	0 … 65 m/s	±0.2 m/s
		Wind direction	0 … 360°	<2°
Particulate matter sensor	Outdoor	PM2.5	0 … 1000 µg/m3	±5 µg/m3	[43]

gives an overview of the installed indoor and outdoor sensors, including the measured variables, measurement ranges, and stated accuracies. All measurement data were recorded at one-minute time intervals, except for the signals of the window contact switches, which were typically logged every 15 min or on an event basis only. The concept and results of the long-term monitoring are described in detail in [29].

Outdoor air quality data provide the outdoor context for interpreting indoor exposure conditions. Ambient PM_2.5 concentrations at the two school sites were of comparable magnitude during the study period, with slightly higher levels observed at the DV-associated site. The relationship between indoor and ambient PM_2.5 concentrations, including the influence of window opening behavior, is analyzed in [29].

2.8. Data Analysis and Statistical Inference

All analyses described in this section were pre-specified before data inspection. Data processing, visualization and statistical analyses were performed primarily in Matlab R2023b (The MathWorks Inc., Natick, MA, USA) [44], using the Statistics and Machine Learning Toolbox version 23.2 (R2023b) [45]. Linear mixed-effects models (LMM) were fitted with the Matlab function fitlme [46]; fixed-effect tests were based on Type III ANOVA with degrees of freedom estimated via the Satterthwaite approximation. Residual diagnostics (histograms, Q–Q plots, residuals-versus-fitted plots) were inspected for all LMM outcomes (sCRP, sIgA, WHO-5). Residual normality was assessed using the Jarque–Bera test, and deviations from normality were considered acceptable given the robustness of LMMs to moderate non-normality. Statistical analyses of the WURSS-K questionnaire were conducted in jamovi version 2.6 (The jamovi project, Sydney, Australia) [47]. Across all analyses, group differences between the two ventilation groups or within-group temporal changes were evaluated at a significance level of α = 0.05. For numerical stability in reporting, p-values smaller than 1 × 10⁻¹⁶ were set to 0.

2.8.1. Statistical Analysis of WURSS-K

For the analysis of the continuous WURSS-K questionnaire outcomes, average (arithmetic mean across the number of reported sick days) and maximum values were calculated for each participant and each parameter. The distributional properties of these variables were assessed separately for the two study groups using the Shapiro–Wilk test. Normally distributed variables were compared between groups using Welch’s t-test, which does not assume equality of variances. When the assumption of normality was violated in either group, the non-parametric Mann–Whitney U test was employed. Unadjusted p-values were reported throughout. No correction for multiple testing (e.g., false discovery rate) was applied because the WURSS-K subscales represent correlated facets of a single underlying construct rather than independent outcomes. For the assessment of infection related outcomes, odds ratio and absolute risk reduction were calculated.

2.8.2. Statistical Analysis of Salivary Biomarkers

Salivary biomarker concentrations (sIgA, sCRP) were log₁₀-transformed prior to analyses due to their positively skewed distributions. Statistical inference was performed on the transformed scale, whereas descriptive statistics were reported on the raw scale for interpretability. In the primary analysis, an LMM with a random intercept for participants was fitted, including group (WA, DV), time (T1–T3), and their interaction as fixed effects. Post hoc pairwise contrasts were computed to compare groups at each time point and to assess within-group temporal changes. Resulting p-values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate. Effect estimates were reported as log₁₀ differences and back-transformed into ratios for interpretation on the original scale.

Additionally, a sensitivity analysis was conducted for sIgA to assess potential confounding by indoor environmental conditions (CO₂, PM_2.5, VOC, temperature and relative humidity). Classroom-level covariates were derived from the long-term monitoring data. For each classroom and each sampling time point (T1–T3), mean values of the five indoor air parameters were calculated across all occupied school hours within the two weeks preceding saliva collection. A two-week reference period was chosen to reflect the typical environmental exposure prior to biomarker collection while minimizing the influence of short-term fluctuations. CO₂ concentrations were scaled per 100 ppm and mean-centered, while all other covariates were standardized (z-scores). These classroom-level values were assigned identically to all children within the same classroom for each time point. Simplified additive LMMs (random intercept for participants) without interaction terms were used to avoid overparameterization given the limited number of classrooms and to ensure stable estimation of covariate effects.

2.8.3. Statistical Analysis of WHO-5

An LMM with group (WA, DV), time (T1–T3), and their interaction as fixed effects, and a random intercept for participants, was applied to assess potential differences in WHO-5 well-being scores.

2.8.4. Statistical Analysis of Comfort-13 Questionnaire

The comfort questionnaire was self-developed for this study and is not a validated instrument. Therefore, results are presented descriptively only. In the discussion, descriptive trends were compared qualitatively with indoor environmental measurements obtained from the long-term monitoring data.

2.8.5. Statistical Analysis of Long-Term Monitoring Data

Long-term IEQ data collected during school hours were summarized using boxplots to illustrate central tendencies and variability across ventilation concepts. As recommended in [3,48], data were aggregated on a school hour basis (45-min lessons, excluding breaks). For group-level comparisons, school-wide mean values aggregated across classrooms within each ventilation concept were used. To ensure comparability between both groups, analyses were restricted to periods during which the decentralized AHUs operated correctly (586 out of 720 school hours). Further details on the data processing are provided in [29].

3. Results

3.1. Study Participants and Baseline Characteristics

At the start of the study, 109 of 174 children from the three ventilation groups and eight selected classrooms participated (

Table 3. Distribution of study participants to ventilation groups. * Group CV was excluded from further analyses.

	Number of Schoolchildren	Number of Study Participants	Participation Rate in %
Group WA	87	64	74
Group DV	46	34	74
Group CV *	41	11	27
Total for all 3 groups	174	109	63
Total for WA & DV	133	98	74

). However, participation rates varied substantially between groups: only 11 children were enrolled in the Central Ventilation (CV) group (27% participation), while both the window airing and decentralized ventilation groups reached 74% participation. Subsequent analyses focus on groups WA and DV, comprising a total of six classrooms. Figure 3 presents the study flow chart according to STROBE recommendations. Among both groups, 98 of 133 pupils participated, with 64 in group WA and 34 in group DV. During the study period, no participant formally dropped out. One child provided only the first saliva sample but continued completing the questionnaires. In both groups some children were absent from school at individual measurement time points (e.g., due to illness) and therefore did not participate in the corresponding in-school assessments. In addition, one comfort questionnaire in group DV was lost, and a small number of saliva samples could not be analyzed due to insufficient sample volume. All 98 participating children were retained in the study, resulting in a final sample of 98 participants. All available data were included in the analyses (Appendix A 

Table A1. Number of available outcome measurements for groups WA and DV.

	Group WA					Group DV
Time Point	WHO-5	Comfort	sCRP	sIgA	WURSS-K	WHO-5	Comfort	sCRP	sIgA	WURSS-K
T1	60	60	50	57		31	30	29	29
T2	57	57	52	53		33	33	31	29
T3	58	58	48	55		34	34	31	29
Continuous					37					17

), and missing observations were treated as outcome-level missingness.

Baseline characteristics of the participating children are summarized in

Table 4. Baseline characteristics of the study population for groups WA and DV.

Parameter		Unit	Total	WA	DV
Gender	Female		45	26	19
	Male		53	38	15
Age	Mean	Years	8.6	8.8	8.1
	Median	Years	8	9	8
	Standard deviation	Years	0.8	0.8	0.6
	Minimum	Years	7	8	7
	Maximum	Years	11	11	9
Asthma	Proportion	%	5.2	4.8	5.9
Allergies	Proportion	%	17.3	23.4	5.9
Smokers in the household	Proportion	%	11.3	11.1	11.8
Pet(s) in the household	Proportion	%	52.0	48.4	58.8
Runny nose in the past 12 months	Proportion “never”	%	9.3	11.1	5.9
	Proportion “sometimes”	%	82.5	81.0	85.3
	Proportion “regularly”	%	8.2	7.9	8.8
Cough in the past 12 months	Proportion “never”	%	12.2	17.2	2.9
	Proportion “sometimes”	%	81.6	78.1	88.2
	Proportion “regularly”	%	6.1	4.7	8.8
Sore throat in the past 12 months	Proportion “never”	%	27.6	26.6	29.4
	Proportion “sometimes”	%	68.4	71.9	61.8
	Proportion “regularly”	%	4.1	1.6	8.8

. No substantial differences in gender or age distribution were observed between groups WA and DV. The proportion of girls and boys was approximately balanced across both groups, with 40.6% girls in group WA and 55.9% in group DV. The mean age at baseline was 8.6 years (standard deviation SD = 0.8), ranging from 7 to 11 years. The prevalence of asthma, exposure to household smoking, pet ownership, and recent respiratory symptoms (runny nose, cough, sore throat) was comparable between groups. A higher prevalence of allergies was observed in group WA, reflecting natural variation in classroom composition. Overall, no indications of systematic baseline imbalances were found.

3.2. Long-Term Monitoring and Classroom-Level Covariates

A detailed analysis of the long-term monitoring data has been reported in [29]. The key values relevant for the present analysis are summarized below. Figure 4 shows boxplots of school-wide mean values of the five indoor air parameters during school hours for both groups, restricted to periods with correct decentralized AHU operation (586 out of 720 school hours). The two groups differed significantly across all five indoor air parameters. Compared with the window-airing group WA, the decentralized ventilation group DV exhibited significantly lower CO₂ and PM_2.5 concentrations, as well as higher indoor air temperatures. Relative humidity was lower in group DV, whereas VOC concentrations were higher. Overall, group DV maintained a higher proportion of comfortable indoor air conditions during school hours, primarily due to higher indoor air temperatures. For the school hours, threshold exceedances and underruns were evaluated for temperature, relative humidity, and CO₂ concentrations. According to the standard DIN EN 16798-1, a minimum operative temperature of 20 °C is recommended for IEQ category II classrooms [35], which was used to assess indoor air temperature [29]. The FGK-Status-Report 58 recommends a minimum relative humidity of 40% in view of potential disease transmission via the respiratory tract [49]. A CO₂ concentration below 1000 ppm is considered hygienically safe [2,3]. Indoor air temperatures below 20 °C occurred in 21.2% of the monitored hours in WA but in only 8.2% in DV. Relative humidity below 40% was common in both groups but occurred more frequently in DV (62.1%) than in WA (46.1%). CO₂ concentrations exceeded 1000 ppm for 26.6% of school hours in WA, compared with only 1.7% in DV. These differences were accompanied by marked differences in window opening behavior. Manual window airing averaged 28 min per school hour in WA and 13 min in DV.

Classroom-level environmental covariates for the statistical analyses were calculated as described in Section 2.8.2.

Table 5. Classroom-level mean indoor environmental covariates for the two weeks preceding each sampling time point (T1–T3).

Group	Classroom	Time Point	CO2 (ppm)	PM2.5 (µg/m3)	VOC (%)	T (°C)	r.h. (%)
WA	1	T1	946	4.6	10.2	21.3	43.9
WA	1	T2	810	5.7	11.5	21.3	31.3
WA	1	T3	938	4.3	9.4	22.5	36.6
WA	2	T1	786	2.4	9.5	20.2	43.5
WA	2	T2	845	4.8	12.8	20.1	32.9
WA	2	T3	858	3.6	8.8	21.5	37.8
WA	3	T1	932	4.0	15.1	19.3	47.7
WA	3	T2	1076	5.8	13.5	18.0	39.9
WA	3	T3	977	4.6	12.4	20.7	41.3
WA	4	T1	960	3.5	16.1	21.8	43.4
WA	4	T2	918	4.8	10.5	20.2	34.9
WA	4	T3	902	4.0	13.0	21.3	40.0
DV	5	T1	764	1.9	18.1	21.3	40.6
DV	5	T2	692	2.3	20.1	21.3	32.0
DV	5	T3	702	2.1	18.1	21.1	37.4
DV	6	T1	849	1.5	15.6	20.2	44.5
DV	6	T2	787	2.3	21.2	19.7	36.3
DV	6	T3	785	1.2	17.2	21.7	37.5

summarizes the mean values of indoor CO₂, PM_2.5 and VOC concentrations, indoor air temperature and relative humidity for each classroom and sampling time point (T1–T3).

3.3. Primary Outcome: Infection Dynamics Measured with WURSS-K

Over the study period, differences between the two ventilation groups WA and DV were observed, using the WURSS-K questionnaire.

Table 6. WURSS-K–descriptive statistics for groups WA and DV. (SD: standard deviation).

WURSS-K		Group
Parameter	Descriptive Statistics	WA	DV
Sick children	Total number	37	17
	Relative proportion in %	57.8	50.0
Sick days	Total number	293	89
	Maximum number per child	24	24
	Minimum number per child	1	2
	Average number per child	7.9	5.2
Illness episodes	Total number	60	25
	Maximum number per child	3	3
	Average number per child	1.6	1.5
	Average number sick days per episode	4.9	3.6
Global total score (items 1–14)	Average (±SD)	9.2 (4.6)	7.8 (5.7)
	Maximum	31	36
Global illness severity (item 1)	Average (±SD)	1.2 (0.5)	1.0 (0.5)
	Maximum	3	3
Symptom severity (items 2–7)	Average (±SD)	3.9 (1.7)	4.0 (2.7)
	Maximum	17	16
Functional impact (items 8–14)	Average (±SD)	4.2 (3.2)	2.9 (3.1)
	Maximum	19	17

provides an overview of the descriptive statistics. In group WA, 57.8% of children reported feeling ill on at least one day during the study period, whereas in group DV, 50% of children reported illness on at least two days. On average, children in group WA experienced symptoms for a total of 7.9 days across the study period, compared to 5.2 days in group DV. The average number of illness episodes (sequence of consecutive days with reported symptoms) was 1.6 in group WA and 1.5 in group DV. The average duration of each illness episode was 4.9 days in group WA and 3.6 days in group DV, corresponding to a 1.3-day shorter duration in the ventilated classrooms.

The following WURSS-K indices were analyzed as described in Section 2.5: The global total score (range: 0–42; higher values indicate stronger symptoms and higher functional impact), illness severity (0–3), symptom severity (0–18), and functional impact (0–21). Average global total score was approximately 15% lower in Group DV than in group WA. Average global illness severity rating was about 17% lower in group DV compared to WA. Average symptom severity scores were nearly identical between groups, with group DV showing a marginal increase of approximately 3%. Average functional impact score was around 31% lower in group DV than in group WA.

Group comparisons were conducted using Welch’s t-tests for normally distributed data and Mann–Whitney U tests for non-normal data.

Table 7. WURSS-K–inferential results for differences between groups WA and DV.

Parameter	Test	Statistic	p		Effect Size
Average global total score	Mann–Whitney U	U = 237	0.152	Rank biserial correlation	−0.2464
Maximum global total score	Mann–Whitney U	U = 207	0.046 *	Rank biserial correlation	−0.3418
Average global illness severity	Welch’s t	t(30.8) = 1.51	0.140	Cohens d	0.4448
Maximum global illness severity	Mann–Whitney U	U = 187	0.013 *	Rank biserial correlation	−0.4054
Average symptom severity	Mann–Whitney U	U = 285	0.589	Rank biserial correlation	−0.0938
Maximum symptom severity	Mann–Whitney U	U = 220	0.077	Rank biserial correlation	−0.3005
Average functional impact	Mann–Whitney U	U = 224	0.092	Rank biserial correlation	−0.2893
Maximum functional impact	Mann–Whitney U	U = 206	0.042 *	Rank biserial correlation	−0.3466
Sick days	Mann–Whitney U	U = 202	0.036 *	Rank biserial correlation	−0.3577
Illness episodes	Mann–Whitney U	U = 278	0.454	Rank biserial correlation	−0.1161

Note: Asterisks indicate significance levels: p < 0.05 (*).

summarizes the inferential results, including test statistics, p-values, and effect sizes. No significant group difference was observed in the average global total score (U = 237, p = 0.152, r = −0.25). In contrast, the maximum global total score differed significantly between groups (U = 207, p = 0.046, r = −0.34), with higher scores in group WA. For global illness severity, no significant difference was found in average values (t(30.8) = 1.51, p = 0.140, d = 0.44), while the maximum global illness severity was significantly higher in group WA (U = 187, p = 0.013, r = −0.41). No statistically significant differences were observed for the symptom severity (U = 285, p = 0.589, r = −0.09 for average values; U = 220, p = 0.077, r = −0.30 for maximum values). The average functional impact did not differ significantly between groups (U = 224, p = 0.092, r = −0.29), while the maximum functional impact was significantly higher in group WA (U = 206, p = 0.042, r = −0.35). Regarding illness duration, the number of sick days was significantly higher in group WA compared to group DV (U = 202, p = 0.036, r = −0.36). The number of illness episodes did not differ significantly between groups (U = 278, p = 0.454, r = −0.12). Across all WURSS-K parameters, the direction of effects consistently indicated higher illness burden, functional impairment, and illness duration in group WA compared to group DV.

Table 8. WURSS-K–calculation of infection rate, absolute risk reduction and odds ratio for groups WA and DV.

	Infected (n)	Not Infected (n)	Total (n)	Infection Rate	Odds of Infection
Group WA	37	27	64	37/64 × 100 = 57.81%	37/27 = 1.37
Group DV	17	17	34	17/34 × 100 = 50.0%	17/17 = 1
Absolute risk reduction (ARR)				57.81 − 50% = 7.81%
Odds ratio (OR)					1.37/1 = 1.37
95% confidence interval (CI)					0.59–3.16

summarizes infection related outcomes. The infection rates derived from the WURSS-K questionnaire indicated that 57.8% of children in classrooms without mechanical ventilation (WA) reported at least one episode of illness during the study period, compared to 50.0% in classrooms with ventilation (DV). This corresponds to an absolute risk reduction (ARR) of 7.8% in favor of the ventilated classrooms. The odds ratio (OR) of 1.37 (95% CI [0.59, 3.16]) indicates 37% higher odds of reporting an infection in the control group WA compared to group DV. Although the odds of infection were higher among children in non-ventilated classrooms, this difference was not statistically significant according to Fisher’s exact test (p = 0.525).

3.4. Secondary Outcome Parameters

3.4.1. Salivary C-Reactive Protein

Salivary C-reactive protein was analyzed across the three measurement points (T1–T3) to assess potential group differences between classrooms without and with mechanical ventilation. Descriptive statistics of the raw concentration data are presented in

Table 9. Salivary C-reactive protein (sCRP)–descriptive statistics at the three time points for groups WA and DV. (SD: standard deviation).

Group	Time Point	Number	sCRP in pg/mL
			Mean	SD	Median	Minimum	Maximum
WA	T1	50	359.04	835.77	77.94	2.08	5470.19
	T2	52	421.13	1400.39	88.26	0.66	9997.15
	T3	48	523.22	1428.66	50.73	2.56	8819.04
DV	T1	29	209.83	483.91	52.15	16.73	2388.11
	T2	31	251.13	487.17	108.00	19.72	2587.41
	T3	31	163.98	285.46	69.24	3.36	1271.81

Note: Values are presented on the original scale (pg/mL).

. At baseline (T1), mean sCRP concentrations were 359.0 pg/mL in group WA and 209.8 pg/mL in group DV. Across all time points, sCRP levels exhibited substantial variability within both groups, with high standard deviations (SD) indicating pronounced inter-individual differences. In group WA, mean concentrations slightly increased from T1 (359.0 pg/mL) to T3 (523.2 pg/mL), whereas in group DV, sCRP values decreased from T1 (209.8 pg/mL) to T3 (164.0 pg/mL). However, these fluctuations did not follow a consistent pattern.

Statistical methods are described in Section 2.8.2. The linear mixed-effects model revealed no significant main effect of group (F(1, 154.4) = 0.02, p = 0.879), indicating that average sCRP levels did not differ between the ventilation conditions (

Table 10. Results of linear mixed-effects model (random intercept for participant) for salivary C-reactive protein after log₁₀ transformation.

Effect	F(df1, df2)	p
Group	0.02 (1, 154.4)	0.879
Time	0.44 (2, 157.3)	0.647
Group × Time	0.69 (2, 151.2)	0.504

). Likewise, there was no significant main effect of time (F(2, 157.3) = 0.44, p = 0.647), suggesting that mean sCRP concentrations remained stable across T1–T3. The group × time interaction was also not significant (F(2, 151.2) = 0.69, p = 0.504), indicating that temporal patterns of sCRP concentrations were similar in both groups. Overall, salivary sCRP concentrations showed no systematic differences between groups or over time.

3.4.2. Salivary Immunoglobulin A

Descriptive statistics of salivary immunoglobulin A concentrations across the three measurement time points (T1–T3) are summarized in

Table 11. Salivary immunoglobulin A (sIgA)–descriptive statistics at the three time points for groups WA and DV. (SD: standard deviation).

Group	Time Point	Number	sIgA in µg/mL
			Mean	SD	Median	Minimum	Maximum
WA	T1	57	154.27	103.46	138.01	8.96	526.91
	T2	53	166.61	114.68	128.31	23.39	474.09
	T3	55	117.30	80.16	105.78	5.93	315.70
DV	T1	29	72.24	45.56	74.21	8.11	186.00
	T2	29	73.98	47.93	60.63	4.20	189.85
	T3	29	72.26	46.23	70.18	1.75	195.77

Note: Values are presented on the original scale (µg/mL).

. At baseline (T1), the mean sIgA concentration in group WA without mechanical ventilation was 154.3 µg/mL, whereas the mean in group DV with mechanical ventilation was 72.2 µg/mL. Across all time points, children in group DV consistently exhibited lower mean sIgA levels compared with group WA. In group WA, the mean concentration increased slightly from T1 to T2 (166.6 µg/mL) but subsequently declined to 117.3 µg/mL at T3, whereas sIgA levels in group DV remained relatively stable throughout the study period.

Table 12. Results of linear mixed-effects model (random intercept for participant) for salivary immunoglobulin A after log₁₀ transformation.

Effect	F(df1, df2)	p
Group	14.89 (1, 164.0)	<0.001 ***
Time	8.01 (2, 161.5)	<0.001 ***
Group × Time	2.60 (2, 160.4)	0.078

Note: Asterisks indicate significance levels: p < 0.001 (***).

summarizes the results of the LMM. The analysis revealed significant main effects of group (F(1, 164.0) = 14.89, p < 0.001) and time (F(2, 161.5) = 8.01, p < 0.001), while the interaction between group and time did not reach statistical significance (F(2, 160.4) = 2.60, p = 0.078).

Post hoc pairwise comparisons, corrected for multiple testing using the Benjamini–Hochberg false discovery rate (FDR), indicated that mean sIgA concentrations were significantly lower in group DV compared to group WA at all measurement points (−52.1% at T1, −59.2% at T2, and −39.1% at T3; all p_FDR < 0.05;

Table 13. Post hoc pairwise comparisons of salivary IgA concentrations between and within groups (FDR-adjusted p-values). Estimates (β) refer to log₁₀-transformed data; ratios and percent changes were back-transformed (10^β).

Contrast	Comparison Type	β (log10)	Ratio	Change (%)	pFDR
DV vs. WA @ T1	Between groups	−0.319	0.48	−52.1	<0.001 ***
DV vs. WA @ T2	Between groups	−0.390	0.41	−59.2	<0.001 ***
DV vs. WA @ T3	Between groups	−0.215	0.61	−39.1	0.010 *
WA T2–T1	Within group	+0.045	1.11	+11.0	0.638
WA T3–T1	Within group	−0.131	0.74	−26.0	0.016 *
DV T2–T1	Within group	−0.025	0.94	−5.5	0.692
DV T3–T1	Within group	−0.027	0.94	−6.0	0.692

Note: Asterisks indicate significance levels: p < 0.05 (*) and p < 0.001 (***).

). Within-group comparisons revealed a significant decrease in sIgA levels in group WA from T1 to T3 (−26.0%, p_FDR = 0.016), whereas no significant temporal changes were observed in group DV (Table 13).

Figure 5 visualizes the mean sIgA concentrations (±standard error (SE)) for both ventilation groups across the three measurement time points. Children in group DV consistently showed lower sIgA mean values than those in group WA. Significant between-group differences, adjusted using the FDR method, were present at all time points (p_FDR < 0.05). Within-group comparisons confirmed the observed decline in sIgA levels from T1 to T3 in group WA (p_FDR = 0.016), while no significant change occurred in group DV.

To examine whether classroom environmental conditions contributed to differences in sIgA concentrations, each indoor environmental parameter was added individually as a classroom-level covariate into a simplified additive LMM (see Section 2.8.2 and Section 3.2). Fixed-effect estimates and back-transformed ratios are summarized in

Table 14. Fixed-effect estimates and back-transformed effects of linear mixed-effects models for salivary sIgA with individual classroom-level covariates. Estimates (β) refer to log₁₀-transformed data; ratios and percent changes were back-transformed (10^β). CO₂,_100,c: CO₂ scaled per 100 ppm and mean-centered; z subscript: covariates standardized to z-scores; n.s.: not significant.

Model	Effect	β (log10)	Ratio (10β)	Change (%)	p	Interpretation
CO2,100,c	Group (DV vs. WA)	−0.176	0.67	−33.39	0.035 *	sIgA lower in DV
	CO2 (+100 ppm)	0.090	1.23	23.08	0.003 **	Positive CO2 association
PM2.5,z	Group (DV vs. WA)	−0.049	0.89	−10.60	0.704	group difference n.s.
	PM2.5 (per 1 SD)	0.148	1.41	40.72	0.015 *	Positive PM2.5 association
VOCz	Group (DV vs. WA)	−0.287	0.52	−48.32	0.003 **	sIgA lower in DV
	VOC (per 1 SD)	−0.014	0.97	−3.25	0.696	n.s.
Tz	Group (DV vs. WA)	−0.306	0.49	−50.57	<0.001 ***	sIgA lower in DV
	T (per 1 SD)	−0.023	0.95	−5.18	0.445	n.s.
r.h.z	Group (DV vs. WA)	−0.292	0.51	−48.99	<0.001 ***	sIgA lower in DV
	r.h. (per 1 SD)	0.061	1.15	15.05	0.241	n.s.

Note: Asterisks indicate significance levels: p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***).

.

Including CO₂ as a covariate revealed a significant positive association with sIgA (β = 0.090, p = 0.003), corresponding to an estimated +23% increase per +100 ppm CO₂. The group effect (DV vs. WA) remained significant after adjustment, indicating that the between-group difference could not be explained solely by mean CO₂ levels. Among the remaining environmental parameters, only PM_2.5 showed a significant positive association with sIgA (β = 0.15, p = 0.015), corresponding to an estimated +41% increase per 1 SD of PM_2.5 (SD = 1.43 µg/m³). When PM_2.5 was included, the group effect was no longer significant, suggesting that classroom-level PM_2.5 partly accounted for the observed differences between ventilation types. Temperature, relative humidity, and VOC concentrations were not significantly associated with sIgA (all p > 0.2).

3.4.3. WHO-5 Well-Being Index

Descriptive WHO-5 scores for both groups across the three time points are provided in Appendix A 

Table A2. WHO-5 well-being index–descriptive statistics of WHO-5 score at the three time points for groups WA and DV. (SD: standard deviation).

Group	Time Point	Number	WHO-5 Score
			Mean	SD	Median	Minimum	Maximum
WA	T1	60	59.4	21.8	64	4.0	96.0
	T2	57	55.6	22.1	60	0	96
	T3	58	57.2	19.6	60	8	96
DV	T1	30	52.8	16.3	58	16.0	76.0
	T2	33	66.9	18.0	68	0	88
	T3	34	59.4	17.1	64	20	84

. Across the study period, mean WHO-5 scores in group WA remained relatively stable (55.6–59.4), while group DV showed a moderate increase at T2 (66.9), compared to T1 (52.8) and T3 (59.4). A linear mixed-effects model (Section 2.8.3) revealed no significant main effect of group (F(1, 207.9) = 2.49, p = 0.116) and no significant main effect of time (F(2, 180.0) = 1.31, p = 0.272). The group × time interaction reached statistical significance (F(2, 178.2) = 8.30, p < 0.001); however, post hoc comparisons showed no significant differences between groups at any individual measurement point after FDR correction (all p_FDR = 0.116).

3.4.4. Comfort-13 Questionnaire

Table A3. Comfort-13 questionnaire–descriptive statistics of the 13 comfort items at the three time points for groups WA and DV. (SD: standard deviation).

		Group WA						Group DV
Time Point	Item	Number	Mean	SD	Median	Minimum	Maximum	Number	Mean	SD	Median	Minimum	Maximum
t1	How awake do you feel today?	60	1.8	1.2	2	0	4	31	1.8	1.0	2	0	4
	How well did you sleep last night?	60	1.5	1.2	1	0	4	31	1.2	0.9	1	0	4
	How do you feel today?	60	1.0	0.8	1	0	4	30	0.7	0.8	0.5	0	3
	How does the air in the room feel to you?	60	1.6	1.0	2	0	4	31	1.4	1.0	2	0	4
	Do you feel any wind in the room?	59	1.1	0.6	1	0	3	30	0.2	0.6	0	0	3
	Do you have dry eyes?	60	0.4	0.7	0	0	3	31	0.4	0.5	0	0	2
	Do you have dry lips?	58	1.3	0.9	1	0	4	31	1.0	0.8	1	0	3
	How does the temperature in the room feel to you?	58	−1.2	1.2	−1	−3	1	31	0.6	1.1	1	−2	3
	How do your hands feel?	60	−0.7	1.5	−1	−3	2	31	0.7	1.4	1	−3	3
	How do your feet feel?	59	−0.6	1.4	−1	−3	3	31	−0.1	1.4	0	−3	2
	How does your head feel?	59	0.4	1.3	0	−3	3	31	1.0	1.0	1	−1	3
	How do you find the volume in the room?	59	1.0	0.6	1	0	2	30	1.1	0.3	1	1	2
	Are there certain noises or sounds that bother you?	55	0.4	0.5	0	0	1	31	0.4	0.5	0	0	1
t2	How awake do you feel today?	57	1.8	1.1	2	0	4	33	1.4	0.9	1	0	4
	How well did you sleep last night?	56	1.5	1.2	1	0	4	33	1.0	0.9	1	0	4
	How do you feel today?	57	1.2	1.1	1	0	4	33	1.1	0.8	1	0	3
	How does the air in the room feel to you?	57	1.8	1.2	2	0	4	33	1.2	1.0	1	0	3
	Do you feel any wind in the room?	57	0.5	0.6	0	0	2	33	0.3	0.4	0	0	1
	Do you have dry eyes?	55	0.5	0.7	0	0	3	33	0.5	0.7	0	0	2
	Do you have dry lips?	56	1.4	1.0	1	0	3	33	1.1	1.0	1	0	3
	How does the temperature in the room feel to you?	56	0.3	1.7	0	−3	3	32	0.8	0.9	1	−1	3
	How do your hands feel?	57	0.5	1.6	1	−3	3	33	0.9	1.1	1	−2	3
	How do your feet feel?	57	0.2	1.4	0	−3	3	32	0.2	1.2	0	−2	2
	How does your head feel?	55	0.8	1.3	1	−3	3	33	1.1	0.9	1	0	3
	How do you find the volume in the room?	55	1.3	0.7	1	0	2	33	1.0	0.3	1	0	2
	Are there certain noises or sounds that bother you?	55	0.6	0.5	1	0	1	33	0.6	0.5	1	0	1
t3	How awake do you feel today?	58	1.8	1.1	2	0	4	34	1.9	0.9	2	1	4
	How well did you sleep last night?	58	1.2	1.1	1	0	4	34	1.3	1.0	1	0	4
	How do you feel today?	58	1.4	1.0	1	0	4	34	1.4	1.5	1	0	9
	How does the air in the room feel to you?	58	1.6	1.1	1	0	4	34	1.6	0.7	2	0	3
	Do you feel any wind in the room?	58	1.1	0.7	1	0	3	34	0.8	0.6	1	0	3
	Do you have dry eyes?	58	0.5	0.9	0	0	3	34	0.6	0.5	1	0	1
	Do you have dry lips?	58	1.2	0.9	1	0	3	34	0.9	0.6	1	0	3
	How does the temperature in the room feel to you?	58	−0.9	1.6	−1	−3	3	34	0.3	1.0	0	−1	3
	How do your hands feel?	58	−0.6	1.7	−1	−3	3	34	0.4	1.4	1	−3	3
	How do your feet feel?	58	−0.8	1.4	0	−3	2	34	0.1	1.3	0	−3	3
	How does your head feel?	58	0.7	1.3	0	−3	3	34	0.9	0.8	1	−1	3
	How do you find the volume in the room?	57	0.7	0.7	1	0	2	34	0.7	0.5	1	0	2
	Are there certain noises or sounds that bother you?	58	0.7	0.4	1	0	1	34	0.6	0.5	1	0	1

(Appendix A) summarizes descriptive statistics for all 13 comfort items. Across all three measurement time points (T1–T3), response patterns were similar between the two groups, with most mean item scores falling within the comfortable/neutral (0) to moderate (2) range of their respective scales.

Children in both groups reported generally favorable to moderate levels of subjective alertness, sleep quality and general mood, with mean values ranging from 0.7 to 1.9 across all time points. Ratings of dry eyes were very low in both groups (0.4–0.6), while reports of dry lips were slightly higher (0.9–1.4). Some descriptive differences emerged between ventilation groups. Children in group DV reported slightly less draught sensation (e.g., T1: mean = 0.2 vs. 1.1 in WA) and slightly better perceived air quality (e.g., T2: mean = 1.2 vs. 1.8 in WA). Thermal sensation tended to be warmer in group DV and cooler in WA (e.g., T1: mean = +0.6 vs. −1.2 in WA). Children in group WA more frequently reported having cooler hands and feet.

Acoustic comfort was similar between groups. Mean loudness ratings ranged from 0.7 to 1.3 (scale: 0–2), and disturbing noises from 0.4 to 0.7 (scale: 0–1). When disturbing sounds were reported, they primarily reflected classroom-related sounds such as “other children, talking, chairs, or falling objects”. In group WA, “heating” was noted as a disturbing noise several times at T2 and T3; this was not reported in group DV. At T3, mentions of “rustling”, “wind,” and once “air conditioner” occurred in group WA, although no air-conditioning or ventilation units were installed in these classrooms. No noises attributable to decentralized AHUs were reported in group DV.

During data collection, windows were open in the WA classrooms at T1 due to elevated CO₂ levels indicated by the CO₂ traffic-light sensors, and at T3 a window was open in group DV. These conditions may have influenced immediate thermal and draught perceptions.

4. Discussion

4.1. Discussion of Results

The WURSS-K outcomes showed a clear and consistent pattern indicating lower illness burden, functional impairment, and illness duration in group DV compared to group WA. Notably, several parameters–including maximum global total score, maximum global illness severity, maximum functional impact, and sick days–showed statistically significant group differences, while all remaining parameters demonstrated effects pointing in the same direction. This convergence across multiple symptom domains suggests that decentralized ventilation may be associated with a measurably milder subjective illness experience in school-aged children. The consistency of these findings is further supported by the epidemiological indicators, including a lower infection rate (absolute risk reduction of 7.8%) in group DV. Together, the symptom-based and epidemiological measures strengthen the interpretation that mechanical ventilation may contribute to reduced perceived illness severity. Although the WURSS-K was defined as the primary outcome of the study, the analyses of its individual symptom domains must be interpreted with caution. The instrument as a whole served as the predefined primary endpoint, whereas the evaluation of multiple correlated subparameters is inherently exploratory. Nonetheless, the consistent direction of effects across all domains–including several statistically significant group differences–supports the robustness of the overall pattern favoring decentralized ventilation.

To complement the symptom-based self-assessment, the salivary biomarkers sCRP and sIgA were examined, as they offer objective measures of inflammatory responses and mucosal immune activity. sCRP levels did not differ significantly between ventilation groups and showed no systematic variation over time. The absence of group, time, or interaction effects indicates that low-grade inflammatory activity, as reflected by sCRP, was not meaningfully influenced by the indoor environmental conditions. Given the low concentrations typical for healthy children and the high inter-individual variability of sCRP, these null findings are consistent with its limited sensitivity to moderate environmental fluctuations.

In contrast, the primary analysis of sIgA revealed clear and consistent group differences: children in mechanically ventilated classrooms showed significantly lower sIgA concentrations across all three time points, and these differences remained robust after FDR correction. This pattern may reflect reduced antigenic stimulation in mechanically ventilated rooms, which exhibited markedly lower CO₂ and PM_2.5 levels in the long-term monitoring. Reduced exposure to airborne pollutants and potentially infectious particles could plausibly diminish mucosal immune activation and thereby lower sIgA secretion.

The sensitivity analysis supports this interpretation. When indoor-environmental parameters were added individually as covariates, most did not affect the group differences. Both PM_2.5 and CO₂ were positively associated with sIgA, indicating that higher exposures predicted higher mucosal immune activity. However, only the inclusion of PM_2.5 eliminated the group effect, suggesting that particulate exposure is a key environmental factor linking the ventilation concept with sIgA levels. The substantially lower PM_2.5 concentrations in mechanically ventilated classrooms therefore offer a coherent explanation for the lower sIgA levels observed in group DV.

These findings align with previous research showing associations between particulate matter exposure and respiratory symptoms in children. Prior work using the WURSS-K has linked higher PM levels to increased respiratory symptom scores in school-aged populations [16]. The present study extends this evidence by integrating repeated symptom assessments with salivary biomarker data and continuous exposure monitoring, providing a comprehensive and temporally resolved perspective under real-world school conditions.

To complement these immune-related findings with a broader psychosocial perspective, general subjective well-being was assessed using the WHO-5 questionnaire. No significant differences in WHO-5 well-being scores were observed between classrooms with and without mechanical ventilation, nor did well-being change systematically over time. Overall, these findings suggest that the ventilation concept did not markedly influence general subjective well-being as measured by the WHO-5. Because the WHO-5 reflects broad psychosocial well-being rather than immediate indoor-environmental perceptions, a more targeted assessment of comfort was needed to examine potential links between ventilation strategy and subjective experience.

Subjective comfort ratings obtained with the Comfort-13 questionnaire were largely consistent with the objectively monitored indoor environmental conditions. Long-term monitoring showed that classrooms with decentralized ventilation exhibited higher indoor air temperatures, lower CO₂ concentrations, and substantially reduced PM_2.5 levels, while relative humidity tended to be lower (frequently remaining below the recommended minimum level of 40%) and VOC levels were higher than in window-aired classrooms. These environmental patterns aligned well with several descriptive trends in the self-reported comfort data. Children in group DV reported warmer thermal impressions and slightly less draught sensation, consistent with the higher temperatures and reduced window-opening durations. Conversely, participants in group WA more frequently reported cooler sensations and cold extremities, which correspond to the objectively lower temperatures and the more intensive window airing that also explains their higher draught ratings. Perceived air quality was descriptively better in group DV, mirroring its markedly lower CO₂ levels. The elevated VOC concentrations in DV, however, did not translate into poorer subjective air-quality ratings, suggesting that concentrations remained below perceptual thresholds or were overshadowed by other comfort cues such as warmth and low CO₂. Acoustic comfort was comparable across groups, with no indications that decentralized ventilation units contributed to noise. Reported disturbances primarily arose from regular classroom activity in both groups or heating noise in group WA, which reflects the operation of radiators. Finally, isolated procedural factors–such as open windows during questionnaire completion at some time points–may have transiently influenced thermal or draught perceptions, but do not alter the overall pattern of results and reflect typical conditions in everyday school life.

4.2. Strenghts and Novelty

A key strength of this study lies in its integrative, real-world design, which combines repeated health assessments and salivary biomarkers with continuous long-term monitoring of IEQ in occupied classrooms. Whereas previous school studies typically focused either on IEQ parameters or on health-related outcomes, the present work integrates these domains within a single analytical framework. This multidimensional approach allows subjective symptoms, mucosal immune markers, and objective exposure data to be interpreted jointly rather than in isolation. A major advantage of this design is the ability to incorporate classroom-level environmental covariates into the statistical models and to explore potential exposure-related mechanisms. The sensitivity analyses showed that the association between ventilation concept and mucosal immune activity was largely driven by PM_2.5. After adjustment for PM_2.5, the group effect on sIgA disappeared, highlighting particulate exposure as a key environmental factor and shifting the interpretation toward exposure-related pathways. To our knowledge, this PM_2.5-mediated pathway linking ventilation concept with mucosal immune responses has rarely been demonstrated in school-based field studies under real operating conditions. Ecological validity is further strengthened by conducting the study during regular school operation and the winter infection season. The convergence of objective IEQ measurements with children’s comfort perceptions enhances the internal coherence of the results and supports their practical interpretability. Finally, the use of a validated symptom instrument (WURSS-K), repeated measurements over time, and mixed-effects modelling accounting for inter-individual variability increases statistical robustness. Taken together, these strengths demonstrate that the present study advances existing research by providing integrated, health-centered evidence on the relationships between ventilation concepts, indoor environmental quality, comfort, and respiratory health in schools.

4.3. Constraints and Limitations

Several methodological constraints should be considered when interpreting the findings. The study followed an observational, non-randomized design determined by existing school infrastructure, which may introduce unmeasured confounding; therefore, causal inferences cannot be drawn. The overall sample size was modest, reducing power for detecting small effects. Despite the modest sample size, the longitudinal repeated-measures design and the integration of continuous IEQ monitoring increased statistical efficiency and allowed detection of effects of moderate magnitude under real-world school conditions. Classrooms with central mechanical ventilation could not be included due to insufficient participation and inconsistent ventilation operation, resulting in unequal group sizes and an imbalanced study design. Inter-individual variability in symptom perception and immune responsiveness is an inherent characteristic of pediatric research and cannot be fully controlled. In the present study, repeated measurements and linear mixed-effects models with subject-specific random intercepts were used for the biomarker analyses to account for between-child variability. Nevertheless, residual heterogeneity between children may have contributed to unexplained variance and limits conclusions at the individual level. Several outcomes relied on self-reported measures from children and may therefore be affected by subjective interpretation and situational influences. This also applies to the WURSS-K symptom scores, which, despite showing a consistent pattern across groups, are inherently variable and may not fully capture day-to-day symptom fluctuations. Subjective outcomes may also be influenced by expectation bias, as blinding of participants and teachers was not possible. Indoor environmental parameters were measured only within the school setting, whereas exposures and activities outside school were not captured, although they may also influence respiratory symptoms and immune responses.

Finally, the generalizability of the findings is influenced by climatic and operational context. The study was conducted in a temperate climate during the heating season, which strongly affects window-opening behavior and thermal comfort. Accordingly, the applicability of manual window airing may differ in warmer or colder climates. In contrast, the decentralized mechanical ventilation system provided a norm-compliant outdoor air volume flow in accordance with DIN EN 16798-1 (IEQ category II) for the investigated classroom sizes. One classroom was operated with a constant air volume flow, while the second was switched to demand-controlled operation based on CO₂ concentration during the study, without compromising compliance with the targeted ventilation requirements. This indicates that the underlying mechanical ventilation concept is, in principle, transferable to other climatic regions when appropriately designed and operated. Moreover, the integrative methodological approach applied in this study is independent of climate and can be readily transferred to investigations conducted under different climatic and regulatory conditions.

5. Conclusions

This study indicates that decentralized mechanical ventilation may provide health- and comfort-related benefits for schoolchildren. The primary outcome, the WURSS-K, showed a consistent pattern of milder subjective illness experiences in mechanically ventilated classrooms. Mucosal immune activity also differed between ventilation concepts, with lower sIgA levels observed in mechanically ventilated classrooms. Although causal inferences cannot be drawn from this observational study, converging evidence from the primary analyses, sensitivity analyses and long-term monitoring of IEQ parameters indicates that ventilation-related differences in particulate exposure (PM_2.5) may influence mucosal immune responses. Children’s comfort perceptions aligned well with the objectively measured indoor environmental conditions. Mechanically ventilated classrooms exhibited warmer and more stable thermal environments during the heating period, significantly lower CO₂ and PM_2.5 concentrations and slightly more favorable thermal and draught-related impressions, although relative humidity was frequently below recommended levels. Importantly, children did not report any acoustic disadvantages related to decentralized mechanical ventilation. These findings suggest that decentralized mechanical ventilation can enhance health and perceived comfort while supporting favorable indoor environmental quality.

A particular strength of this study lies in combining medical outcomes with continuous long-term monitoring of indoor environmental quality under real-world school conditions, an approach that remains uncommon in school-based ventilation research. This integrative methodological approach enabled the inclusion of environmental classroom-level covariates and allowed subjective perceptions, immune markers, and objective exposure data to be interpreted jointly. Additional strengths include the use of a validated symptom instrument and the ecological validity afforded by measurements during regular school operation. Comparable levels of detail and multidimensional integration have rarely been applied in school-based research. Future studies should employ larger cohorts and ensure balanced participation across all ventilation groups to further clarify the relationships between ventilation concept, indoor environmental quality, and child health.

From a practical perspective, the findings support decentralized mechanical ventilation as a robust option to improve indoor environmental quality, air hygiene, and perceived comfort in classrooms, particularly during the heating season when window airing alone may be insufficient. At the same time, potential trade-offs related to investment costs, energy use, maintenance requirements, and material-related emissions should be considered during planning and operation, emphasizing the importance of appropriate system design and operation. Operational aspects of mechanical ventilation in the same study setting have been discussed in more detail in a previous monitoring-focused publication [29]. Overall, ventilation strategies should be selected contextually, balancing air hygiene, comfort, energy efficiency, and operational constraints, and evaluated using integrated approaches that consider both environmental conditions and health-related outcomes.