Effectiveness of a structured protocol for the prevention of acute respiratory tract infections in daycare centers: A quasi-experimental study

Abstract

BACKGROUND

Acute respiratory tract infections (ARTIs) are a major cause of pediatric morbidity. Daycare centers (DCCs) are high-risk environments where close interpersonal contact facilitates rapid viral transmission. Effective and practical preventive strategies are therefore essential to reduce infection-related burden in these settings.

AIM

To determine the incidence of ARTIs in DCCs following implementation of a structured protocol designed to mitigate respiratory virus transmission.

METHODS

A cluster study was conducted in which two centers were centers to either intervention or control groups via lottery. Although allocation was randomized, the study was not designed as a randomized controlled trial; therefore, potential bias and confounding were considered. Participants were followed for 25 weeks. The primary outcome was physician-diagnosed ARTIs per 100 child-weeks. Incidence rate ratios (IRRs) with 95%CIs were estimated using Poisson regression with a log link and person-time offset, adjusting for covariates under the intention-to-treat principle. Results cautious interpretation Due to limited clusters, results require cautious interpretation. Statistical significance was defined as P < 0.05.

RESULTS

A total of 223 children (149 intervention, 74 control) across 2 clusters were enrolled. By week 25, the cumulative proportion of ARTIs was 18.1% in the intervention group compared with 60.8% in the control group. The incidence rate of ARTIs was lower in the intervention group, with adjusted rates of 1.09 per 100 child-weeks vs 13.00 per 100 child-weeks. The adjusted IRR was 0.14 (95%CI: 0.09-0.22; P < 0.001), suggesting an observed difference in incidence between groups under the study conditions.

CONCLUSION

Protocol implementation was associated with a lower observed incidence of ARTIs under the study conditions. However, interpretation is constrained by the study design, the small number of clusters, and potential sources of bias. Larger, well-designed randomized studies are required to further evaluate this association in daycare settings.

Key Words: Child daycare centers; Common cold; Infectious disease transmission; Prevention; Respiratory tract infection

Core Tip: Acute respiratory tract infections remain a major health concern among children in day care centers. Practical prevention strategies are therefore needed. This study observed an association between the proposed protocol and a reduction in infection incidence under the study conditions. These findings may have implications for informing infection prevention practices in similar day care settings.

INTRODUCTION

Acute respiratory tract infections (ARTIs) remain one of the most common causes of morbidity among young children worldwide, particularly in environments where close contact is unavoidable, such as daycare centers, early childhood development centers, and preschools[1,2]. These infections are primarily transmitted via respiratory droplets, direct contact, and contaminated surfaces, with young children being especially susceptible due to immunologic immaturity and frequent close-contact behaviors. Consequently, ARTIs continue to represent a significant global public health burden and a leading cause of pediatric healthcare utilization and hospitalization.

In Thailand, respiratory tract infections constitute a major cause of pediatric illness and hospital admission, particularly among children younger than five years[3-5]. Data from Naresuan University Hospital, a tertiary care center, indicate that lower respiratory tract infections are among the most frequent infectious diagnoses in pediatric patients, with viral pathogens accounting for the majority of cases[3]. Pneumonia, bronchiolitis, and bronchitis remain the predominant clinical presentations, contributing substantially to disease burden and healthcare resource utilization[3]. These findings underscore the persistent impact of respiratory viral infections on child health at both local and national levels.

During the coronavirus disease 2019 (COVID-19) pandemic, the widespread implementation of nonpharmaceutical interventions (NPIs), including mask use, hand hygiene, and environmental cleaning, was associated with a marked reduction in the circulation of common respiratory viruses in children[6-8]. However, the evidence is heterogeneous, reflecting differences in intervention intensity, adherence, timing, and population context[9-11]. Following the relaxation of these measures and the reopening of schools and daycare centers, a resurgence of respiratory viral infections has been observed in many settings, often as clustered outbreaks[6-8]. These observations suggest that while NPIs can be effective, their impact may not be sustained without structured, context-specific implementation strategies[12-14].

Despite this, a critical knowledge gap remains regarding the effectiveness of cluster-level implementation of standardized, multi-component respiratory infection prevention protocols with adherence monitoring in early childhood education settings, particularly in low- and middle-income countries (LMICs). Existing studies have largely focus on hospital-based interventions or household-level practices, with limited evidence from real-world daycare environments where sustained close contact among children facilitates transmission. Furthermore, prior studies often evaluate single-component interventions rather than integrated protocols, and few have systematically assessed implementation fidelity or adherence over time, which are key determinants of effectiveness.

The present study addresses this gap by evaluating a structured, multi-component prevention protocol that integrates mask use, hand hygiene, and routine environmental cleaning, alongside behavioral reinforcement and adherence monitoring. This approach is hypothesized to be more effective than single or uncoordinated interventions by simultaneously targeting multiple transmission pathways and promoting sustained compliance in high-contact settings. Therefore, this study aimed to assess whether the implementation of this protocol was associated with a reduced incidence of ARTIs among children aged 2-6 years in daycare settings, providing scalable evidence to inform infection prevention strategies in LMICs.

MATERIALS AND METHODS

Patient and public involvement

This study involved children aged 2-6 years who were enrolled at the Tha Thong Subdistrict Early Childhood Development Center and the Hua Ro Subdistrict Early Childhood Development Center in Mueang District, Phitsanulok Province, Thailand. Neither patients nor members of the public were involved in the development of the study design, the implementation of the intervention, or the preparation of the manuscript. Children participated solely through study enrollment and age-appropriate data collection procedures (Figure 1).

Figure 1  Study flow diagram of cluster allocation, follow-up, and analysis.

Study design and setting

This study was conducted as a comparative case study with quantitative analysis in two early childhood development centers in Mueang District, Phitsanulok Province, Thailand (Tha Thong and Hua Ro Subdistrict centers), between May 15 and December 15, 2025. The two centers constituted the study clusters, with one assigned to the intervention group and the other to the control group. Given the inclusion of only two clusters, this design is subject to inherent methodological constraints. In particular, cluster-level confounding cannot be resolved, and observed between-group differences may reflect underlying structural or contextual variation rather than the intervention alone.

Trial design

This research is registered with the Thai Clinical Trials Registry (TCTR) under the identification number TCTR20250422001, and the registered title “Effectiveness of a Protocol for the Prevention of Acute Respiratory Tract infection in Daycare Centers”.

The study employed a cluster quasi-experimental, parallel-group design with random allocation of clusters (lottery-based assignment), in which each early childhood development center constituted a single cluster. All eligible children within the same center received the same study condition to minimize contamination between participants. Although clusters were randomly assigned, the study does not meet the methodological criteria of a cluster randomized trial due to the inclusion of only two clusters; therefore, effect estimates should be interpreted with caution.

Two clusters-the Tha Thong Subdistrict Early Childhood Development Center and the Hua Ro Subdistrict Early Childhood Development Center, both located in Mueang District, Phitsanulok Province, Thailand-were included in the study. Cluster selection was performed using a multistage sampling approach, followed by random allocation to either the intervention or control group. Given that only one cluster was included per study arm, intervention effects cannot be fully distinguished from underlying cluster-specific characteristics.

The intervention cluster (Tha Thong Subdistrict Early Childhood Development Center) implemented a structured protocol (Table 1) aimed at preventing ARTI. The protocol was applied uniformly to all enrolled children, teachers, and relevant staff within the center. The control cluster (Hua Ro Subdistrict Early Childhood Development Center) continued routine practices without exposure to the intervention and was observed prospectively over the same period.

Table 1 Structured multimodal protocol for preventing respiratory tract infections in childcare centers.

Time	Activity	Responsible personnel
07:00 a.m.-08:00 a.m.	Teachers conduct health screening for children aged 2-6 years, including temperature measurement and assessment for skin lesions, rashes, and oral abnormalities (e.g., erythematous throat, tonsillar swelling, oral ulcers), to identify potential respiratory infections. Children with suspected infection are referred to a designated healthcare facility. Those without signs of infection are instructed to wear a surgical face mask, perform hand hygiene using alcohol-based hand sanitizer, and attend classes as usual	Duty teacher
08:00 a.m.-11:00 a.m.	During class, teachers assess surgical face masks for contamination and replace them if soiled. Ensure that children wear surgical face masks at all times, except during meals and sleep. Require hand hygiene using alcohol-based hand sanitizer after activity changes or class periods. After restroom use, require handwashing with soap and water and followed by hand drying	Classroom teacher
11:00 a.m.-01:00 p.m.	Remove surgical face masks before meals and perform hand hygiene before and after eating. Surgical face masks are not required during sleep. After restroom use, require handwashing with soap and water followed by hand drying	Classroom teacher
01:00 p.m.-03:00 p.m.	During class, teachers assess surgical face masks for contamination and replace them if necessary. Ensure continuous surgical face masks use except during meals and sleep. Require hand hygiene after activity transitions and after restroom use	Classroom teacher
03:00 p.m.-04:00 p.m.	Clean and disinfect high-touch surfaces (e.g., doorknobs, shared toys) using disinfectants, dishwashing liquid, or soapy water	Housekeeping staff
After 04:00 p.m.	Indoor areas: Classroom floors and walls, and shared toys are cleaned once daily using cleaning agents or disinfectants; Outdoor areas: Cafeteria, playgrounds equipment, shared play equipment, and restrooms are cleaned once daily using cleaning agents or disinfectants	Housekeeping staff

All design features-including intervention delivery, follow-up schedule, outcome assessment, and data collection-were implemented at the cluster level, whereas outcomes were measured at the individual child level. This cluster-based approach was selected to reflect real-world daycare settings and to reduce cross-exposure between intervention and control groups. However, it limits the ability to distinguish individual-level effects from cluster-level influences. The intervention and follow-up period lasted 6 months, with assessments conducted at 12 scheduled visits.

Eligibility criteria for clusters

Early childhood development centers were eligible for inclusion as clusters if they met the following criteria: (1) Located within Mueang District, Phitsanulok Province, Thailand; (2) Provided routine daycare services for children aged 2-6 years; (3) Had stable enrollment of children and staff during the study period; and (4) Agreed to participate in the study and permit implementation or observation of the study procedures throughout the 6-month intervention period. Centers were excluded if they were implementing other infection prevention interventions or programs that could interfere with the study protocol during the study period.

Intervention

The intervention was implemented at both the cluster and individual participant levels. At the cluster level, the intervention was applied to the entire early childhood development center assigned to the intervention group. This included center-wide implementation of a structured protocol designed to limit the transmission of respiratory tract infections, along with training and engagement of teachers and support staff to ensure consistent application of the protocol within daily routines. The intervention protocol comprised multiple components, including mask use, hand hygiene, and routine environmental cleaning, targeting different transmission pathways.

At the individual participant level, children enrolled in the intervention cluster and their caregivers received education and guidance on proper mask use and hand hygiene practices. Prior to participation, each child underwent a brief physical assessment to identify any contraindications to mask use. Mask use was adapted to be age-appropriate, particularly for children aged 2-3 years, based on developmental capability and safety considerations. In this age group, masks use were used under close adult supervision to ensure proper fit without causing breathing discomfort and were limited to periods when children were awake and calm. Masks were not used during sleep or meals, and caregivers were instructed to monitor for signs of intolerance (e.g., breathing difficulty, distress, or frequent removal).

Adherence to the intervention protocol was reinforced through regular follow-up visits, during which compliance and implementation challenges were monitored by the research team. Quantitative measures of implementation fidelity were assessed throughout the intervention period. Mask adherence was approximately 60%-75% among children aged 2-3 years (n = 41), compared with 90%-98% among children older than 3 years. Hand hygiene compliance ranged from 95% to 97% across all age groups, while adherence to environmental cleaning protocols reached 100%.

The control cluster did not receive any intervention at either the cluster or individual level and continued routine practices throughout the study period. Outcomes were monitored in both clusters over the same follow-up period to allow comparison of intervention effects.

Outcomes

Outcome measures in this study primarily pertained to the individual participant level, with additional descriptive assessment at the cluster level. At the individual level, outcomes included the occurrence and frequency of ARTIs among children aged 2-6 years during the 6-month intervention period. These outcomes were assessed prospectively for each child and served as the primary basis for comparison between the intervention and control clusters. An ARTI episode was defined using a standardized symptom-based case definition: The presence of at least two respiratory symptoms (e.g., cough, rhinorrhea, nasal congestion, sore throat, or difficulty breathing), with or without fever (≥ 38.0 ℃), lasting for a minimum duration of 24 hours. Where feasible, symptom reports were supported by clinical assessment by trained personnel; however, laboratory confirmation was not routinely performed. To distinguish recurrent events, a new ARTI episode was defined by a symptom-free interval of at least 7 consecutive days following resolution of the previous episode.

At the cluster level, aggregate outcomes were derived by summarizing individual-level data within each early childhood development center. These included the overall proportion of children experiencing at least one episode of ARTIs and patterns of protocol adherence within each cluster. Cluster-level summaries were used to describe and compare trends between centers rather than as independent primary outcomes. This hierarchical outcome structure reflected the cluster-based study design, in which the intervention was delivered at the center level while health outcomes were measured at the individual child level.

Harms

Potential harms were predefined as any discomfort or difficulties related to implementation of the respiratory infection prevention protocol, particularly surgical mask use and hand hygiene. Harms were assessed systematically throughout the 6-month intervention period via direct observation and structured inquiry during 12 scheduled follow-up visits. Teachers and caregivers reported any protocol-related issues, which were recorded using standardized forms. Children were screened for contraindications to surgical mask use at baseline. No serious adverse events were observed during the study.

Study size estimation

Sample size estimation was performed to determine the theoretical minimum number of participants required to detect a difference in the incidence of ARTIs between groups. Assuming an infection rate of 20% in the non-intervention group and 10% in the intervention group, with a two-sided significance level of 0.05, 80% power, and a 1:1 allocation ratio, the required sample size for an individually randomized design was 39 participants per group.

However, this calculation did not account for the cluster-based design, including within-cluster correlation and the associated design effect, and therefore may underestimate the required sample size. Given the inclusion of only two clusters, reliable estimation of the intracluster correlation coefficient (ICC) and appropriate adjustment using standard cluster methods were not feasible. As a result, the sample size should be interpreted as a pragmatic estimate rather than a formally valid calculation for cluster-based inference.

In practice, the final sample size was determined by the available population within the selected early childhood development centers. All eligible children in each center were invited to participate, resulting in a total of 223 participants, which exceeded the minimum estimated sample size. This pragmatic approach reflects real-world constraints, including the limited number of clusters and study sites, rather than a fully powered cluster-randomized design.

Randomisation

Sequence generation: Given the inclusion of only two early childhood development centers, the sampling and allocation procedures were simplified for clarity and transparency. Two early childhood development centers in Mueang District, Phitsanulok Province, Thailand, were selected using simple random sampling. The names of all eligible centers in the district (n = 35) were written on identical slips of paper, placed in a container, and thoroughly mixed. The principal investigator then randomly drew two centers without replacement. The selected centers were Hua Ro Subdistrict Early Childhood Development Center and Tha Thong Subdistrict Early Childhood Development Center.

Cluster allocation to the intervention or control group was determined using a lottery method. Identical sealed papers labeled “intervention” and “control” were prepared and drawn by the principal investigator, with the process documented in a study log to enhance transparency and reproducibility. No stratification or matching procedures was performed due to the limited number of clusters (n = 2). With only one cluster per study arm, random allocation does not ensure baseline comparability between groups. Allocation concealment mechanism: Allocation was conducted at the cluster level, with the early childhood development center serving as the unit of assignment. Consequently, allocation concealment at the individual participant level was not applicable, as children were not randomized independently.

Given the nature of the intervention and the inclusion of only two clusters, allocation concealment was not feasible at either the cluster or individual level. This transparency is necessary as the lack of concealment introduces potential selection and performance bias, which must be considered when interpreting the study’s findings.

Implementation: The random allocation sequence was generated by the research team using a lottery method. The same research team identified eligible centers and assigned clusters to the intervention or control condition.

Individual participants were included through complete enumeration, whereby all eligible children aged 2-6 years enrolled in each selected center were invited to participate based on predefined inclusion criteria. No individual-level random sampling was performed.

Written informed consent was obtained from the parents or legal guardians of all participating children prior to enrollment. Consent was obtained after cluster allocation but before initiation of the intervention and data collection. Approval from the directors of both early childhood development centers was obtained prior to participant recruitment.

Awareness of assignment: Due to the nature of the intervention and the cluster-based design, blinding was not possible. The research team, teachers, caregivers, and parents were aware of group assignment throughout the study period.

Statistical analyses

All statistical analyses were performed using Stata version 19 (StataCorp, College Station, TX, United States). Baseline characteristics were summarized using mean ± SD for continuous variables and frequency with n (%). Between-group comparisons at baseline were conducted using independent t-tests for continuous variables and χ² tests or Fisher’s exact tests for categorical variables, as appropriate; however, given inclusion of only two clusters, these comparisons are descriptive and do not represent formal tests of baseline equivalence. The primary outcome was the incidence rate of ARTIs, defined as the number of infection episodes per child-week of follow-up. Incidence rates were calculated by dividing the total number of observed episodes by the total number of child-weeks in each group and were expressed per 100 child-weeks.

Incidence rates between groups were compared using Poisson regression models with a log link and an offset for person-time. Incidence rate ratios (IRRs) with 95%CIs were reported. Multivariable models were pre-specified to adjust for potential confounders identified a priori, including age, sex, primary caregiver, parental education, and income sufficiency. Parental education was categorized into three levels (lower secondary or below, upper secondary to diploma, and bachelor’s degree or higher) to reduce sparse-data bias and improve model stability.

Given that the intervention was assigned at the cluster level with only two clusters, key assumptions underlying regression-based inference are not fully satisfied. Cluster-level confounding cannot be addressed, and between-cluster variability cannot be reliably estimated; accordingly, approaches such as random-effects models or cluster-robust standard errors at the cluster level were not feasible. While participant-level clustering was used to account for repeated measures within children, this does not capture between-cluster correlation. Regression analyses were therefore conducted for descriptive and exploratory purposes. Estimates are inherently under-identified at the cluster level, regardless of covariate adjustment, and P-values and confidence intervals are reported for completeness but should be interpreted with caution, as they do not provide reliable measures of inferential certainty in this context.

Given the limited number of clusters and potential imbalance in cluster-level characteristics, adjusted estimates may be sensitive to model specification, particularly where substantial differences between crude and adjusted estimates are observed. The ICC was not estimated, as two clusters are insufficient for reliable estimation of between-cluster variance. Formal sensitivity analyses addressing cluster-level correlation were not feasible; however, model-based robustness analyses were performed, including sequential covariate adjustment and comparison of Poisson and negative binomial models.

Harms were summarized descriptively using frequencies and percentages; no formal hypothesis testing was undertaken due to the absence of serious adverse events. For visualization, the weekly cumulative proportion of children experiencing at least one ARTI episode was calculated using all available follow-up data (Figure 2). Adjusted predicted incidence rates over time were estimated from the multivariable Poisson model using marginal predictions and are presented graphically (Figure 3); these should be interpreted as illustrative trends rather than precise estimates.

Figure 2  Weekly cumulative proportion of children with at least one respiratory tract infections episode in the intervention and control groups over the 25-week follow-up period.

Figure 3  Adjusted predicted mean incidence rates of respiratory tract infections in the intervention and control groups over the 25-week follow-up period, estimated from the multivariable Poisson regression model using marginal predictions.

All participants were analyzed according to their assigned cluster (intervention vs control), consistent with the intention-to-treat principle. Follow-up time accrued from enrollment to week 25 or the last completed assessment. Participants lost to follow-up contributed person-time up to their last observation and were right-censored thereafter.

Missing baseline covariate data were minimal (< 5%) and were handled using complete-case analysis under a missing-at-random assumption. No missing data were observed for key baseline variables. Outcome data were missing only due to loss to follow-up (six children in the intervention group and one in the control group, primarily due to relocation). Given the small proportion of missing data and the use of person-time analysis, the impact on estimates is expected to be minimal.

No subgroup analyses were pre-specified due to the limited number of clusters. Additional pre-specified analyses included multivariable-adjusted models to assess the robustness of crude estimates. Given the design constraints, all statistical inferences should be considered hypothesis-generating. All tests were two-sided, with P < 0.05 considered statistically significant.

RESULTS

Two clusters (early childhood development centers) were included in the study and assigned to the intervention and control groups. In the intervention cluster (Tha Thong Subdistrict Early Childhood Development Center), 149 children were enrolled and received the intended intervention according to the study protocol (Table 1). In the control cluster (Hua Ro Subdistrict Early Childhood Development Center), 74 children were enrolled and received usual care.

During the study period, six children in the intervention cluster and one child in the control cluster were lost to follow-up due to relocation of their families to other provinces. These participants contributed follow-up data up to their last completed visit and were censored thereafter in the analysis. No clusters were excluded after allocation. Consequently, 149 children in the intervention group and 74 children in the control group were included in the analysis of the primary outcome. (Figure.1)

The mean age of patients in the intervention group was 3.5 ± 1.1 years and 2.9 ± 0.5 years in the control group (P < 0.001). Male participants accounted for 74 children (49.7%) in the intervention group and 38 children (51.4%) in the control group, respectively (P = 0.887). A total of 16 participants had underlying diseases, including 13 (8.7%) in the intervention group and 3 (4.1%) in the control group (P = 0.275).

Parents were the primary caregivers for the majority of participants, accounting for 87.9% in the intervention group and 86.5% in the control group (P = 0.788). Grandparents served as the primary caregivers for 8.1% of participants in the intervention group and 10.8% in the control group.

Regarding paternal occupation, the intervention group had the highest proportion of labor workers (39.3%), followed by business or self-employed occupations (28.3%) and government or state enterprise employees (15.9%). Similarly, in the control group, labor workers constituted the largest proportion (45.2%), followed by government or state enterprise employees (31.5%) and business or self-employed occupations (17.8%).

In the intervention group, both fathers and mothers predominantly had educational levels below higher education compared with those in the control group, with statistically significant differences (P < 0.001 and P = 0.021, respectively). Despite this baseline imbalance, the intervention group exhibited a lower incidence of respiratory tract infections than the control group following implementation of the protocol. These findings suggest a potential beneficial effect of the intervention; however, they should be interpreted with caution given the observed differences in parental education, which may act as confounding factors.

The majority of participants in the intervention group reported sufficient household income (123 participants, 84.3%), which was comparable to the control group (60 participants, 81.1%). No statistically significant difference in income sufficiency was observed between the two groups (P = 0.571). Although several baseline characteristics were similar between groups, significant imbalances were noted in participant age and parental educational level (Table 2). These variables were therefore included as prespecified covariates in adjusted analyses.

Table 2 Comparative analysis of baseline demographic, n (%)/mean ± SD.

Characteristics	Missing	Total (n = 233)	Intervention group (n = 149)	Control group (n = 74)	P value
Sex (male)		112 (50.2)	74 (49.7)	38 (51.4)	0.887
Age (years)		3.3 ± 0.1	3.5 ± 1.1	2.9 ± 0.5	< 0.001
BMI (kg/m2)		16.5 ± 3.0	16.4 ± 3.0	16.7 ± 3.0	0.410
Pre-existing conditions		16 (7.2)	13 (8.7)	3 (4.1)	0.275
Primary caregiver					0.788
Parents		195 (87.4)	131 (87.9)	64 (86.5)
Grandparents		20 (9.0)	12 (8.1)	8 (10.8)
Others		8 (3.6)	6 (4.0)	2 (2.7)
Parent’s occupation	5 (2.2)				0.042
Unemployed		5 (2.3)	4 (2.8)	1 (1.4)
Labor		90 (41.3)	57 (39.3)	33 (45.2)
Agriculture		6 (2.8)	5 (3.5)	1 (1.4)
Business		54 (24.8)	41 (28.3)	13 (17.8)
Government		46 (21.1)	23 (15.9)	23 (31.5)
Private sector		15 (6.9)	13 (9.0)	2 (2.7)
Others		2 (0.9)	2 (1.4)	0
Paternal education	9 (4.0)				< 0.001
Primary school		33 (15.4)	23 (16.2)	10 (13.9)
Secondary school		94 (43.9)	77 (54.2)	17 (23.6)
Higher education		87 (40.7)	42 (29.6)	45(62.5)
Maternal education	7 (3.1)				0.021
Primary school		23 (10.7)	15 (10.5)	8 (11.0)
Secondary school		83 (38.4)	66 (46.2)	17 (23.3)
Higher education		110 (51.0)	62 (43.4)	48 (65.8)
Income sufficiency	3 (1.4)				0.571
Sufficient		183 (83.2)	123 (84.3)	60 (81.1)
Insufficient		37 (16.8)	23 (15.8)	14 (18.9)

BMI: Body mass index.

A total of 27 of 149 children (18.1%) in the intervention group and 45 of 74 children (60.8%) in the control group experienced at least one ARTI episode (Supplementary Table 1). Overall, 139 episodes occurred over 3697 person-weeks in the intervention group, compared with 132 episodes over 1842 person-weeks in the control group. The crude incidence rates were 3.76 per 100 person-weeks and 7.17 per 100 person-weeks in the intervention and control groups, respectively, corresponding to a crude IRR of 0.52 (95%CI: 0.31-0.88; P = 0.015).

After adjustment for prespecified covariates (sex, age, primary caregiver, parental educational level, and income adequacy), the adjusted IRR was 0.14 (95%CI: 0.09-0.22; P < 0.001) (Table 3). Sequential modeling demonstrated that adjustment for age substantially reduced the IRR from 0.52 to 0.13, while further adjustment for parental education yielded an IRR of 0.15; inclusion of additional covariates resulted in a final IRR of 0.14. These findings indicate that baseline imbalances in age and parental education accounted for much of the difference between crude and adjusted estimates (Supplementary Table 2).

Table 3 Incidence rate of respiratory tract infections and incidence rate ratios during follow-up, median (interquartile range).

Group	Children (n)	Child-weeks with ARTI (n)	Episodes per child	Observed child-weeks	Incidence rate (per 100 child-weeks)	Crude IRR (95%CI)	P value	Adjusted IRR (95%CI)1	P value
Control	74	132	1 (0-2)	1842	7.17	Reference		Reference
Intervention	149	139	0 (0-0)	3697	3.76	0.52 (0.31-0.88)	0.015	0.14 (0.09-0.22)	< 0.001

¹Adjusted for age, sex, primary caregiver, father’s education, mother’s education, and income sufficiency.

IRR: Incidence rate ratios; ARTI: Acute respiratory tract infections.

To assess model assumptions, overdispersion in the Poisson model was evaluated, and a sensitivity analysis using negative binomial regression was performed. Robust standard errors clustered at the participant level were applied to account for repeated weekly observations. The adjusted negative binomial model yielded results nearly identical to those of the adjusted Poisson model (IRR = 0.14, 95%CI: 0.09-0.22; P < 0.001), with an overdispersion parameter close to zero (alpha < 0.001), indicating no meaningful overdispersion (Supplementary Table 3).

Sequentially adjusted models are presented in Supplementary Table 2, and sensitivity analyses using alternative count-data models are detailed in Supplementary Table 3. These analyses demonstrated a consistent protective direction of effect across different specifications. However, given the small number of clusters, all regression-based estimates should be interpreted with caution.

Cumulative proportion curves showed early divergence and sustained separation between groups (Figure 2). Adjusted predicted incidence rates derived from the multivariable model (Figure 3) indicated that by week 25, the incidence was 1.09 per 100 childweeks in the intervention group compared with 13.00 per 100 childweeks in the control group.

DISCUSSION

Respiratory tract infections remain a major public health concern in pediatric populations, particularly among younger children who are more susceptible to severe disease than older age groups. The burden is further amplified in crowded environments such as daycare centers, where close and sustained contact facilitates pathogen transmission and increases infection incidence. Previous studies have consistently demonstrated that children attending daycare centers experience substantially higher rates of respiratory tract infections-up to two to four times greater than those cared for at home[15-17]. This disparity is largely attributable to the intensity and duration of interpersonal interactions inherent in group-care settings. In response, we developed and implemented a structured preventive protocol tailored to daycare environments.

The findings of this study indicate a lower observed incidence of respiratory tract infections in the intervention group. However, given the study design, these findings should be interpreted as associative rather than causal. When contextualized within the existing literature, the results highlight important differences across study settings and implementation approaches[18,19]. Prior investigations conducted in household settings, such as those by Cowling et al[20] and Simmerman et al[21], did not demonstrate significant reductions in infection rates with combined mask use and hand hygiene. In contrast, the present study was conducted in daycare settings, characterized by higher contact intensity and transmission risk, which may partly explain the observed differences.

Differences in adherence and behavioral reinforcement may also contribute to variability in observed effects. Studies reporting limited or non-significant findings often involved suboptimal adherence or less structured implementation[22,23], whereas the present study incorporated a predefined protocol with prospective adherence monitoring. Adherence was generally high, particularly for hand hygiene and environmental cleaning, and among older children for mask use. Nevertheless, the relationship between adherence and infection outcomes cannot be established within the current design. While adherence may plausibly act as a mediator, this cannot be formally evaluated. Future studies should adopt standardized approaches to measuring adherence and further examine its potential mediating role.

Comparisons across studies are further complicated by differences in intervention structure, particularly between single-component and bundled approaches. Many prior studies evaluated individual measures, whereas the present study implemented a multi-component intervention combining mask use, hand hygiene, and environmental cleaning. While such bundled approaches may better reflect real-world practice, they limit the ability to isolate the contribution of individual components. The observed findings therefore likely reflect the combined influence of these measures. Mechanistically, mask use may reduce exposure to respiratory pathogens, hand hygiene may reduce contact-mediated transmission, and environmental cleaning may decrease surface contamination. However, the relative contributions of these components cannot be disentangled within the current design. Future studies using factorial or component-level approaches are warranted.

Recent post-COVID-19 systematic reviews in school and daycare settings have reported variable effectiveness of non-pharmaceutical interventions, often influenced by contextual factors such as ventilation, adherence, and baseline transmission intensity[24]. These findings underscore that intervention effectiveness is context-dependent rather than uniform across settings. Collectively, the heterogeneity observed across studies likely reflects differences in setting (household vs daycare), adherence, and intervention design (single vs bundled), rather than a consistent presence or absence of effect.

Importantly, the primary contribution of this study lies not in establishing efficacy per se, but in demonstrating the feasibility and structured implementation of a multi-component prevention protocol in a real-world daycare context. The use of predefined procedures, adherence monitoring, and coordinated implementation may provide a practical framework for infection prevention in similar settings.

Several alternative explanations should be considered. Unmeasured structural differences between centers-such as ventilation, crowding, and staff-to-child ratios-may have contributed to the observed differences. Differential healthcare-seeking or reporting behaviors between groups may also have influenced outcome detection. Although standardized procedures were applied, subtle differences in outcome ascertainment may persist. In addition, a Hawthorne effect arising from more intensive monitoring in the intervention cluster cannot be excluded. Given the small number of clusters, effect size inflation is also plausible.

In this study, the marked difference between crude and adjusted estimates (IRR 0.52 vs 0.14) indicates that the findings are highly sensitive to model specification. This suggests substantial model dependence; therefore, the adjusted results should be interpreted with caution as exploratory rather than confirmatory. Given the potential for residual confounding and the limited ability of covariate adjustment to fully address cluster-level imbalance, these estimates are best viewed as hypothesis-generating, and further studies with more robust designs are needed to confirm the observed associations.

Strengths

This study has several strengths. First, it was conducted in real-world daycare settings, enhancing its practical relevance. However, internal validity is constrained by the study design, particularly the inclusion of only two clusters, which limits the ability to control for cluster-level confounding. Consequently, external validity is limited, and the findings should be considered hypothesis-generating rather than broadly generalizable, particularly within similar daycare settings in LMIC contexts. Second, the cluster-based design minimized contamination between intervention and control groups, although the small number of clusters limits statistical precision. Third, adjustment for key sociodemographic covariates reduced, but did not eliminate, residual confounding. Fourth, although the observed effect size was substantial, it should be interpreted as an association rather than evidence of causality. Finally, the protocol consisted of low-cost, non-pharmacological, and scalable interventions, supporting its potential applicability in similar resource-constrained settings.

Limitations

Several limitations should be acknowledged. The quasi-experimental cluster design is inherently susceptible to residual confounding and selection bias. The inclusion of only two daycare centers with unequal cluster sizes limits statistical power and the ability to account for intra-cluster correlation. Outcome assessment relied partly on caregiver or staff reporting, introducing potential measurement bias, while adherence depended on behavioral compliance, which may have varied across participants. The symptom-based definition of ARTIs is pragmatic but introduces potential misclassification in the absence of laboratory confirmation. Differential misclassification between clusters cannot be excluded, particularly if surveillance intensity or reporting practices differed between groups. For example, more intensive monitoring in the intervention cluster may have led to earlier detection or increased reporting awareness, potentially biasing estimates. Conversely, less complete outcome ascertainment in one cluster could result in overestimation of the observed effect. Although standardized procedures were implemented, subtle differences in reporting and outcome assessment may have persisted. Finally, generalizability is limited by context-specific factors, and multiplicity arising from exploratory analyses cannot be excluded. Accordingly, the findings warrant confirmation in larger multi-center cluster-randomized trials with standardized outcome assessment and longer follow-up periods.

CONCLUSION

The implementation of a structured prevention protocol was associated with a substantial reduction in the incidence of ARTI compared with the control group. However, these findings should be interpreted as preliminary require confirmation in larger, well-controlled studies. While the results suggest the potential for successful application in daycare centers, further evidence from multi-center cluster randomized controlled trials is necessary to establish the protocol’s long-term effectiveness and generalizability across diverse real-world settings.

ACKNOWLEDGEMENTS

The authors would like to express their gratitude to Miss Daisy Gonzales of the International Relations Section, Faculty of Medicine, Naresuan University, for her valuable assistance in editing the manuscript.