Indoor Air 2016; 26: 157–167 wileyonlinelibrary.com/journal/ina Printed in Singapore. All rights reserved
© 2015 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd INDOOR AIR doi:10.1111/ina.12202
Ventilation in day care centers and sick leave among nursery children Abstract Several studies have reported poor indoor air quality (IAQ) in day care centers (DCCs), and other studies have shown that children attending them have an increased risk of respiratory and gastrointestinal infections. The aim of this study was to investigate whether there is an association between ventilation in DCCs and sick leave among nursery children. Data on child sick leave within an 11-week period were obtained for 635 children attending 20 DCCs. Ventilation measurements included three proxies of ventilation: air exchange rate (ACR) measured with the decay method, ACR measured by the perfluorocarbon tracer gas (PFT) method, and CO2 concentration measured over a 1-week period. All but two DCCs had balanced mechanical ventilation system, which could explain the low CO2 levels measured. The mean concentration of CO2 was 643 ppm, exceeding 1000 ppm in only one DCC. A statistically significant inverse relationship between the number of sick days and ACR measured with the decay method was found for crude and adjusted analysis, with a 12% decrease in number of sick days per hour increase in ACR measured with the decay method. This study suggests a relationship between sick leave among nursery children and ventilation in DCCs, as measured with the decay method.
B. Kolarik1, Z. Jovanovic Andersen2, T. Ibfelt3, E. Hoj Engelund4, E. Møller1, E. Vaclavik Br€ auner1 1 Department of Construction and Health, Danish Building Research Institute, Aalborg University, Copenhagen, Denmark, 2Department of Public Health, Center for Epidemiology and Screening, Copenhagen University, Copenhagen, Denmark, 3Departments of Infection Control and Clinical Microbiology, Copenhagen University Hospital (Rigshospitalet), Copenhagen, Denmark, 4DHI, Hoersholm, Denmark
Key words: Air change rate; Absenteeism; Child care; Decay; Perfluorocarbon tracer gas; Carbon dioxide.
Barbara Kolarik Department of Construction and Health, Danish Building Research Institute, Aalborg University, A.C. Meyers Vaenge 15, 2450 Copenhagen SV, Denmark Tel.: +45 2910 0334 e-mail: [email protected] Received for review 18 November 2014. Accepted for publication 15 March 2015.
Practical Implications
It is common knowledge that with entering a day care environment, the frequency of respiratory infections for a child increases as a consequence of the shift to a larger viral and bacterial exposure. This increased morbidity in early age is accompanied with work absence of the parent, who either takes care of the ill child and/or becomes infected by the child, which therefore inflicts a large socioeconomic cost. Despite this study’s limitations, which are presented in the discussion section, this study suggests importance of adequate ventilation in day care centers for the health of children. Further work is needed to establish whether indeed there is a relationship between sickness absence and ventilation rate in day care centers.
Introduction
Scandinavian children spend a large percentage of each day in public day care centers (DCCs), and one of the most important health issues is the transfer of infectious diseases. It has been estimated that around 97% of all Danish children aged 3–6 years attend DCCs (Møller, 2014). On average, these children spend 7 h each day in DCCs and 10% of these children spend 8 h or more (Ottosen et al., 2010). Attendance is slightly
lower for the children aged under 2 years, such that around 19% of children under 1 year and 91% of children between 1 and 2 years attend DCCs (Holm, 2014; Møller, 2014). A few previous studies have shown that children who attend DCCs have an increased risk of respiratory and gastrointestinal infections compared to those that stay at home (Nafstad et al., 2004; Nystad et al., 1999). This effect is most pronounced during the first year of life, when their immune systems are not fully developed (Nafstad et al., 2004; Nystad et al., 157
Kolarik et al. 1999). The increased risk of respiratory and gastrointestinal infections among small children is associated with large socioeconomic costs, as children’s infections are often followed by parents’ work absence because they either need to take care of the child or become infected by the child. A Danish study has estimated that a reduction of one child sick leave day per year would be equivalent to a surplus of 445 million DKK (approximately 81.5 million US$) (Pedersen, 2007). Good indoor air quality (IAQ) is important for health, and inadequate ventilation has been identified as a risk factor for asthma, allergy and other problems, including neurodevelopmental disorders (Bornehag et al., 2005; Emenius et al., 2004a,b; H€agerhed-Engman et al., 2006; Øie et al., 1999). However, most indoor climate research to date has focused primarily on homes and office buildings with respect to health in adults (Leyten et al., 2012; Li et al., 2007; Milton et al., 2000; Sundell et al., 2011; Wargocki et al., 2002). Thus, the effect of indoor ventilation on infectious disease and resulting sick leave in children attending DCCs remains largely unexplored, but this association is possible. In Denmark, the Danish Building Regulation (BR10) (Danish Enterprise and Construction Authority, 2010) that was enforced June 30, 2010 by the Danish Ministry of Economic and Business Affairs, stipulates that inhabitable rooms in newly built DCCs must be ventilated by installations comprising of both forced air supply and exhaust. Furthermore, BR10 stipulates that the outdoor air supply rates should not be 1000 ppm, and more than half of the children attending these day cares were exposed to CO2 concentrations exceeding 2000 ppm for a 20min period (Clausen et al., 2009). Finally, the results of that study indicate that high CO2 concentrations, occupant density and children’s sick leave were strongly correlated (Clausen et al., 2009, 2011, 2012). Nursery children were not included in that study, but the reported effects are expected to be stronger among nursery children, as their immune systems are not fully developed (Nafstad et al., 2004). The importance of ventilation has been highlighted in several comprehensive scientific reviews on the effects of ventilation on health (Sundell et al., 2011; Wargocki et al., 2002). These reviews concluded that there was a strong association between ventilation and airway infections as well as increased short-term sick leave 158
among adults working in offices with outdoor exchange rates below 25 l/s per person. These conclusions are further supported by one study showing that the probability of detecting airborne rhinoviruses was positively associated with weekly average CO2 levels in the office (Myatt et al., 2004), as well as another review that concluded there was strong evidence of an association between ventilation and air movements in buildings with the transmission and spread of infectious diseases such as measles, tuberculosis, chickenpox, influenza, smallpox, and SARS (Li et al., 2007). No international or Danish studies have so far investigated whether there is an association between the air change rate and number of sick leave days among children in DCCs due to infectious disease. It is well established that low air exchange rates result in increased concentration of indoor-generated pollutants that are associated with symptoms of ‘sick building’ syndrome and other discomfort and adverse health effects (Sepp€anen et al., 1999; Wargocki et al., 2002) and similar effects applying to bacteria and viruses are possible. The aim of this study was to investigate whether there is an association between ventilation in DCCs and sick leave due to infectious disease among nursery children.
Methods The study design
The study was linked to an already established ‘Health in Day-care Centres’ (SIB) study (SIB, 2014), which aimed to determine the factors responsible for the spread of infectious diseases among small children in Danish nurseries and kindergartens. The SIB project was mainly focused on hygiene, and ventilation was not included. The ventilation measurements were therefore conducted 1 year later. The study design is shown in Figure 1. Results concerning the effects of hygiene on the spread of infectious diseases, including hand hygiene, cleaning methods and products as well as surface types, have previously been reported elsewhere (Ibfelt et al., 2015a). Day care centers
As a part of SIB project (SIB, 2014), a total of 23 DCCs were randomly recruited within the Greater Copenhagen Area, Zealand (17 DCCs), and Nyborg, Funen (6 DCCs) in the fall of 2011. All these DCCs were contacted again 1 year later for agreement to participate in ventilation measurements. One of these was excluded due to missing information on child age, and two did not give participation consent leaving 20 DCCs and 635 attending children included in this study as depicted in Figure 2.
Kolarik et al. Measurements of ventilation and indoor environmental variables
The ventilation measurements were made 1 year after the registration of sick leave, during January 14, 2013 to March 12, 2013 (Figure 1). Three proxies of ventilation were used in the study: (i) the air exchange rate (ACR) measured with the decay method (ii) the ACR measured by perfluorocarbon tracer gas (PFT) method, and (iii) CO2 concentration. The ACR measurements with decay method were conducted in unoccupied rooms either early in the morning, before children were divided into classes or late in the afternoon, after children were collected. All windows and doors were closed tight, after which carbon dioxide was dosed into the room up to a concentration range of 1500–5000 ppm. The decay of the CO2 concentrations was monitored every 30 s for at least 40 min using CARBOCAPÒ CO2 monitors (GMW115; Vaisala, Vantaa, Finland) connected to HOBO U12 -013 data loggers (Onset, Bourne, MA, USA). The windows and doors remained closed during the measurement. A mixing fan was installed to achieve good mixing of the air during the measurement. The 7-day ACR of each DCC was determined using the PFT method (Bergsøe, 1992; Leaderer et al., 1985). The DCCs were treated as single zones and PFT emitters (6–14 per day care), and capillary adsorption tube samplers (7–21 per day care) were deployed. The number of deployed emitters and samplers was dependent on the size of the measured area in each DCC. The amount of tracer adsorbed in the samplers was analyzed using gas chromatography – electron capture detector (GC-ECD), and the ACR was calculated on the basis of the concentration, measured temperature and known emission rates, as well as the volume of the measured part of the building. The PFT measurements were conducted during normal operation of DCCs. Carbon dioxide was measured with CARBOCAPÒ CO2 monitors associated with external input of HOBO data loggers. CARBOCAP CO2 monitors measure CO2 concentrations in the range 0–5000 ppm with an accuracy of (2% of range + 2% of reading) (Vaisala). The mean CO2 concentrations during opening hours (only working days, from 6 am to 5 pm) was calculated and used as proxy of ventilation. The temperature and relative humidity (RH) were measured continuously during 1 week using HOBO U12 -013 data logger. As given by producer, HOBO loggers measures temperature within range 20 to 70°C with accuracy 0.35°C from 0 to 50°C and relative humidity within range 5–95% with accuracy 2.5% (Onset). Similarly to CO2, mean concentrations were calculated for opening hours only. All measurements were carried out in one playroom and one sleeping room in the nursery part of the DCC. As far as possible, sampling was conducted at a height of 1–1.5 m above the floor and not in immediate prox160
imity of windows, heaters, and doors. However, safety of children was of paramount over optimal setup. Furthermore, visual inspections of buildings and interviews with personnel with focus on ventilation type (mechanical, natural, hybrid), type and age of windows and window opening habits were performed. Finally, it was noted whether the smallest children (0–2 years) slept outside in cots/prams during their midday naptime, which is a common practice in Denmark and would reduce indoor exposure time. Sampling and processing of bacteria and virus
The sampling and processing of bacteria and viruses were conducted in winter 2012 as a part of the SIB project (Figure 1), and it have been previously described in detail elsewhere (Ibfelt et al., 2015b). In brief, gastrointestinal viruses were sampled from surfaces at six locations (three in the playroom and three in the changing room/toilet), airway viruses were sampled from surfaces at three locations (all in the playroom) and bacteria were sampled from surfaces at 15 locations (three in the kitchen, seven in a playroom and five in a changing room/toilet). The results of bacteria measurements in the kitchens were excluded from the analysis as ventilation measurements were not conducted here and children do not have regular access to this room. In contrast, changing rooms and toilets are usually directly connected with the playrooms in Danish DCCs and doors are often open; therefore, these measurements were included in the analysis. More information on sampling procedure and analyzed species can be found elsewhere (Ibfelt et al., 2015a). Statistical analysis
We analyzed the association between ventilation and the number of sick days caused by infectious disease among 635 children attending 20 DCCs in Denmark. We used linear mixed model with number of sick days per child as dependent variable and included DCC as a random factor to account for the fact that children attending same institution are correlated. Separate analyses were conducted for three proxies of ventilation as the main explanatory variable: (i) ACR measured with decay method, (ii) ACR measured with PFT method, and (iii) mean CO2 concentration. The effect of each of the three proxies of ventilation on number of sick days was evaluated in several successive steps including variables considered to be associated with sick leave: (i) model 1: crude model with no adjustments and including DCC as a random factor, (ii) model 2: as for model 1 with further adjustment for gender (female vs. male) and age (3 years), and (iii) model 3: as for model 2 and additionally adjusted for sleeping outside (yes, no)
Kolarik et al. Measurements of ventilation and indoor environmental variables
The ventilation measurements were made 1 year after the registration of sick leave, during January 14, 2013 to March 12, 2013 (Figure 1). Three proxies of ventilation were used in the study: (i) the air exchange rate (ACR) measured with the decay method (ii) the ACR measured by perfluorocarbon tracer gas (PFT) method, and (iii) CO2 concentration. The ACR measurements with decay method were conducted in unoccupied rooms either early in the morning, before children were divided into classes or late in the afternoon, after children were collected. All windows and doors were closed tight, after which carbon dioxide was dosed into the room up to a concentration range of 1500–5000 ppm. The decay of the CO2 concentrations was monitored every 30 s for at least 40 min using CARBOCAPÒ CO2 monitors (GMW115; Vaisala, Vantaa, Finland) connected to HOBO U12 -013 data loggers (Onset, Bourne, MA, USA). The windows and doors remained closed during the measurement. A mixing fan was installed to achieve good mixing of the air during the measurement. The 7-day ACR of each DCC was determined using the PFT method (Bergsøe, 1992; Leaderer et al., 1985). The DCCs were treated as single zones and PFT emitters (6–14 per day care), and capillary adsorption tube samplers (7–21 per day care) were deployed. The number of deployed emitters and samplers was dependent on the size of the measured area in each DCC. The amount of tracer adsorbed in the samplers was analyzed using gas chromatography – electron capture detector (GC-ECD), and the ACR was calculated on the basis of the concentration, measured temperature and known emission rates, as well as the volume of the measured part of the building. The PFT measurements were conducted during normal operation of DCCs. Carbon dioxide was measured with CARBOCAPÒ CO2 monitors associated with external input of HOBO data loggers. CARBOCAP CO2 monitors measure CO2 concentrations in the range 0–5000 ppm with an accuracy of (2% of range + 2% of reading) (Vaisala). The mean CO2 concentrations during opening hours (only working days, from 6 am to 5 pm) was calculated and used as proxy of ventilation. The temperature and relative humidity (RH) were measured continuously during 1 week using HOBO U12 -013 data logger. As given by producer, HOBO loggers measures temperature within range 20 to 70°C with accuracy 0.35°C from 0 to 50°C and relative humidity within range 5–95% with accuracy 2.5% (Onset). Similarly to CO2, mean concentrations were calculated for opening hours only. All measurements were carried out in one playroom and one sleeping room in the nursery part of the DCC. As far as possible, sampling was conducted at a height of 1–1.5 m above the floor and not in immediate prox160
imity of windows, heaters, and doors. However, safety of children was of paramount over optimal setup. Furthermore, visual inspections of buildings and interviews with personnel with focus on ventilation type (mechanical, natural, hybrid), type and age of windows and window opening habits were performed. Finally, it was noted whether the smallest children (0–2 years) slept outside in cots/prams during their midday naptime, which is a common practice in Denmark and would reduce indoor exposure time. Sampling and processing of bacteria and virus
The sampling and processing of bacteria and viruses were conducted in winter 2012 as a part of the SIB project (Figure 1), and it have been previously described in detail elsewhere (Ibfelt et al., 2015b). In brief, gastrointestinal viruses were sampled from surfaces at six locations (three in the playroom and three in the changing room/toilet), airway viruses were sampled from surfaces at three locations (all in the playroom) and bacteria were sampled from surfaces at 15 locations (three in the kitchen, seven in a playroom and five in a changing room/toilet). The results of bacteria measurements in the kitchens were excluded from the analysis as ventilation measurements were not conducted here and children do not have regular access to this room. In contrast, changing rooms and toilets are usually directly connected with the playrooms in Danish DCCs and doors are often open; therefore, these measurements were included in the analysis. More information on sampling procedure and analyzed species can be found elsewhere (Ibfelt et al., 2015a). Statistical analysis
We analyzed the association between ventilation and the number of sick days caused by infectious disease among 635 children attending 20 DCCs in Denmark. We used linear mixed model with number of sick days per child as dependent variable and included DCC as a random factor to account for the fact that children attending same institution are correlated. Separate analyses were conducted for three proxies of ventilation as the main explanatory variable: (i) ACR measured with decay method, (ii) ACR measured with PFT method, and (iii) mean CO2 concentration. The effect of each of the three proxies of ventilation on number of sick days was evaluated in several successive steps including variables considered to be associated with sick leave: (i) model 1: crude model with no adjustments and including DCC as a random factor, (ii) model 2: as for model 1 with further adjustment for gender (female vs. male) and age (3 years), and (iii) model 3: as for model 2 and additionally adjusted for sleeping outside (yes, no)
Ventilation and sick leave in day care centers and municipality (Copenhagen area vs. Nyborg) (Figure 3). Additionally, several sensitivity analyses (models 4– 6) were conducted. Model 4 was fit as model 3 and additionally adjusted for variables which may directly affect three ventilation measures: frequency of windows opening (1, 2, and 3 or more times per day), age of windows (newer with double glazing vs. older with single glazing and/or inner sash) and type of windows (wooden vs. plastic). Model 5 was fit as model 4 and additionally adjusted for mean temperature and RH, which are related both to ventilation measures as well as to sick leave. Finally, model 6 was fit as model 4 and additionally adjusted for possible intermediate variables or proxies of sick leave: gastrointestinal or airway virus and nasopharyngeal or intestinal bacteria (Figure 3). We did not have decay measurements in DCC 2 and 8 (Table 2); thus, the analyses of the effect of decay were based on data excluding these DCCs. We repeated therefore the analyses based on a subset of data excluding DCC 2 and 8, in which we explored the effect of the proxies CO2 and PFT. The significance threshold was P < 0.05, and all analyses were performed in SPSS (IBM SPSS Statistics 22 for Windows, Armonk, NY, USA). Within two of the included DCCs, nursery children, and kindergarten children shared the same rooms (in so called family rooms), and for these two DCCs, the older children (above 3 years) were also included in the analyses. The sick leave data showed skewed distribution; therefore, all regression analyses were performed using the natural logarithm of the number of sick days. The association between sick days caused by infectious disease and each of the ventilation proxies was calculated
Crude model
Model 1: Crude model Day-care as a random factor
Main models
Model 2: Fitted as the crude model with adjustment for gender and age
Sensitivity analyses
Model 4 Fitted as Model 3 with further adjustment for frequency of opening windows, age of windows and window type
by exponential transformation of the regression estimate and expressed as a percentage change. Results
This study included 482 nursery children aged 0– 3 years and 153 kindergarten children aged over 3 years. The included DCCs had 1–5 nurseries (or family groups), with 11–34 children in each group (median 12). The total number of groups in each day care (including both nursery and kindergarten groups) ranged from 2 to 7, and the number of children per DCC ranged from 24 to 149. All but two DCCs had balanced mechanical ventilation systems. The description of the study population and the investigated DCCs is presented in Table 1. Results of temperature, RH and CO2 measurements, as well as ventilation measurements for each DCC are presented in Table 2. The median temperature and RH during measuring period were 22°C and 27%, respectively, and the median concentration of CO2 calculated for opening hours only was 579 ppm (mean 643 ppm). The maximum 20 min running average during opening hours was 1132 ppm, ranging from 681 to 2864 ppm. In 9 DCCs, children experienced at least one 20-min period with CO2 concentrations exceeding 1000 ppm and in one DCC with concentration over 2000 ppm (data not shown). The ACR differed between the DCCs (Table 2) and between measurement techniques. There was no correlation between air change rate measured with PFT and decay method (Pearson coefficient 0.082, P = 0.74) nor was PFT significantly correlated with mean CO2 concentration (Pearson’s coefficient 0.233, P = 0.31). However, there was a significant negative correlation between ACR measured with decay method and mean
Model 3 Fitted as Model 2 with further adjustment for sleeping outside and municipality
Model 5 Fitted as Model 4 with further adjustment for mean temperature and relative humidity which are both related to ventilation measures and sick leave
Model 6 Fitted as Model 4 with further adjustment for possible intermediate variables or proxies of sick leave
Fig. 3 The statistical approach
161
Kolarik et al. Table 1 Characteristic of the study population and the investigated DCCs
Characteristic Room type at DCCa Nursery (children 0–3 years) Family rooms (children 0–6 years) Number of children per day care Child gendera Male Female Child age (years)a 3 Sick leave (days)b Sleeping outsidec Yes In crib rooms No Window agec New Old Window constructionc Wooden Plastic Window opening frequencyc Once a day Twice a day ≥Three times a day Ventilation typec Mechanical Natural Municipalityc Great Copenhagen area, Zealand Nyborg, Funen Samples with detected airway or intestinal bacteriad Samples with detected airway or gastro-viruse
N (%)
Median (5th, 95th percentile)
482 (75.9) 153 (24.1) 36 (12, 90) 331 (52.1) 304 (47.9) 0 (0) 220 (34.6) 273 (43.0) 142 (22.4) 3 (0, 13) 4 (20) 13 (65) 3 (15) 15 (75) 5 (25) 11 (55) 9 (45) 9 (45) 9 (45) 2 (10) 18 (90) 2 (10) 15 (75) 5 (25)
Discussion 4 (0, 6) 16 (8, 20)
a
Data per child (635 children included in analyses). Based on total number of sick-leave days of all included children during 11-week period. c Data per DCC (20 day cares included in analyses). d Maximum 24, see Methods for specification. e Maximum 78, see Methods for specification. b
CO2 concentration (Pearson coefficient 0.691, P = 0.001), such that lower CO2 concentrations were measured in DCCs with higher ACR. The highest number of sick days was recorded in DCCs with lowest ACR measured with the decay method, as well as among boys as compared to girls and among smallest children as compared to older children (results not shown). The inverse relationship between the number of sick days recorded in winter 2012 and ACR measured with decay method in winter 2013 was statistically significant in the crude analysis and the two main adjusted models (model 2 and 3; Table 3), with a 12% (95% confidence interval 2–21%) decrease in the number of sick days per hour increase in ACR. When the variables affecting ventilation rates in a room (model 4) and the intermediate variables (model 6) were included into the model, the association 162
was only slightly attenuated, losing however statistical significance, while adjustment for temperature and RH (model 5) resulted in an even stronger and more statistically significant associations (Table 3). The associations between sick leave and the other two proxies of ventilation measured 1 year after sick leave registration were not statistically significant. The results for CO2 concentration measurements point however to the same direction, showing a 2% increase in sick leave (not significant) per 100 ppm increase in CO2 concentration, while results for ACR measurements with PFT are opposite in sign, such that more sick days have been registered in DCCs with higher ACR measured with PFT (not significant; Table 3). In subanalyses excluding DCCs 2 and 8, which were missing in our analyses of decay, we did not see an effect of CO2 or PFT. The estimates were as follows: PFT: 10.9% change with CI: 4.4 and 28.7 and CO2: 1.8% change with CI: 3.3 and 7.1; P = 0.512 (results not shown). None of the proxies of ventilation were significantly correlated with number of bacteria or virus detected on indoor surfaces 1 year before ventilation measurements (results not shown). A non-significant negative relationship was found between ACR measured with decay and gastro-virus as well as airway bacteria and airway virus, while a non-significant positive relationship was found between ACR measured with decay and intestinal bacteria (results not shown).
This study linked data on sick leave among 635 children attending 20 DCCs with ACR and CO2 measurements from these DCCs conducted a year later. Mean concentrations of CO2 were relatively low in this study, and 1000 ppm was only exceeded in one DCC. Air change rates varied among DCCs and between measuring techniques. Results from this study show an association between sick leave among nursery children and ventilation in DCCs measured with decay method. To our knowledge, the only other previous study investigating sick leave was also conducted in Denmark (Clausen et al., 2011) and reported a significant decrease in sick leave among kindergarten children (aged 3–6 years) associated with a decrease in the maximum 20 min average CO2 concentration (Clausen et al., 2011). Although results are not directly comparable with ours as we studied mainly nursery children (0–3 years of age), we report the same trends in the association between sick leave and CO2, showing a 2% increase (not significant) in sick leave per 100 ppm increase in CO2 concentration. In addition, previous results reported in studies in school and office environments corroborate this association (Ervasti et al., 2012; Milton et al., 2000; Shendell et al., 2004; Simons et al., 2010).
Ventilation and sick leave in day care centers Table 2 Temperature, relative humidity, concentration of carbon dioxide and air change rate (ACR) during measuring period
Day care 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Reference
Ventilation (N-natural, M-mechanical)
Temperature, °C Median (5th, 95th percentile)
M M M M M M M M M M N M M M M M M M N M
22.5 (21.2, 24.5) 21.5 (20.2, 22.3) 21.3 (20.3, 22.1) 22.6 (20.8, 24.3) 24.1 (23.5, 25.9) 21.7 (20.9, 22.5) 21.0 (20.1, 22.0) 25.7 (20.9, 27.2) 21.7 (19.8, 23.1) 22.0 (21.0, 22.9) 20.3 (17.9, 21.6) 21.4 (20.7, 22.1) 19.8 (19.1, 20.4) 21.8 (20.8, 22.8) 21.7 (19.1, 22.9) 22.5 (20.6, 24.4) 23.9 (23.2, 24.2) 21.9 (21.2, 22.8) 20.2 (19.6, 21.1) 22.5 (21.4, 23.3) Summer 23°C
© 2015 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd INDOOR AIR doi:10.1111/ina.12202
Ventilation in day care centers and sick leave among nursery children Abstract Several studies have reported poor indoor air quality (IAQ) in day care centers (DCCs), and other studies have shown that children attending them have an increased risk of respiratory and gastrointestinal infections. The aim of this study was to investigate whether there is an association between ventilation in DCCs and sick leave among nursery children. Data on child sick leave within an 11-week period were obtained for 635 children attending 20 DCCs. Ventilation measurements included three proxies of ventilation: air exchange rate (ACR) measured with the decay method, ACR measured by the perfluorocarbon tracer gas (PFT) method, and CO2 concentration measured over a 1-week period. All but two DCCs had balanced mechanical ventilation system, which could explain the low CO2 levels measured. The mean concentration of CO2 was 643 ppm, exceeding 1000 ppm in only one DCC. A statistically significant inverse relationship between the number of sick days and ACR measured with the decay method was found for crude and adjusted analysis, with a 12% decrease in number of sick days per hour increase in ACR measured with the decay method. This study suggests a relationship between sick leave among nursery children and ventilation in DCCs, as measured with the decay method.
B. Kolarik1, Z. Jovanovic Andersen2, T. Ibfelt3, E. Hoj Engelund4, E. Møller1, E. Vaclavik Br€ auner1 1 Department of Construction and Health, Danish Building Research Institute, Aalborg University, Copenhagen, Denmark, 2Department of Public Health, Center for Epidemiology and Screening, Copenhagen University, Copenhagen, Denmark, 3Departments of Infection Control and Clinical Microbiology, Copenhagen University Hospital (Rigshospitalet), Copenhagen, Denmark, 4DHI, Hoersholm, Denmark
Key words: Air change rate; Absenteeism; Child care; Decay; Perfluorocarbon tracer gas; Carbon dioxide.
Barbara Kolarik Department of Construction and Health, Danish Building Research Institute, Aalborg University, A.C. Meyers Vaenge 15, 2450 Copenhagen SV, Denmark Tel.: +45 2910 0334 e-mail: [email protected] Received for review 18 November 2014. Accepted for publication 15 March 2015.
Practical Implications
It is common knowledge that with entering a day care environment, the frequency of respiratory infections for a child increases as a consequence of the shift to a larger viral and bacterial exposure. This increased morbidity in early age is accompanied with work absence of the parent, who either takes care of the ill child and/or becomes infected by the child, which therefore inflicts a large socioeconomic cost. Despite this study’s limitations, which are presented in the discussion section, this study suggests importance of adequate ventilation in day care centers for the health of children. Further work is needed to establish whether indeed there is a relationship between sickness absence and ventilation rate in day care centers.
Introduction
Scandinavian children spend a large percentage of each day in public day care centers (DCCs), and one of the most important health issues is the transfer of infectious diseases. It has been estimated that around 97% of all Danish children aged 3–6 years attend DCCs (Møller, 2014). On average, these children spend 7 h each day in DCCs and 10% of these children spend 8 h or more (Ottosen et al., 2010). Attendance is slightly
lower for the children aged under 2 years, such that around 19% of children under 1 year and 91% of children between 1 and 2 years attend DCCs (Holm, 2014; Møller, 2014). A few previous studies have shown that children who attend DCCs have an increased risk of respiratory and gastrointestinal infections compared to those that stay at home (Nafstad et al., 2004; Nystad et al., 1999). This effect is most pronounced during the first year of life, when their immune systems are not fully developed (Nafstad et al., 2004; Nystad et al., 157
Kolarik et al. 1999). The increased risk of respiratory and gastrointestinal infections among small children is associated with large socioeconomic costs, as children’s infections are often followed by parents’ work absence because they either need to take care of the child or become infected by the child. A Danish study has estimated that a reduction of one child sick leave day per year would be equivalent to a surplus of 445 million DKK (approximately 81.5 million US$) (Pedersen, 2007). Good indoor air quality (IAQ) is important for health, and inadequate ventilation has been identified as a risk factor for asthma, allergy and other problems, including neurodevelopmental disorders (Bornehag et al., 2005; Emenius et al., 2004a,b; H€agerhed-Engman et al., 2006; Øie et al., 1999). However, most indoor climate research to date has focused primarily on homes and office buildings with respect to health in adults (Leyten et al., 2012; Li et al., 2007; Milton et al., 2000; Sundell et al., 2011; Wargocki et al., 2002). Thus, the effect of indoor ventilation on infectious disease and resulting sick leave in children attending DCCs remains largely unexplored, but this association is possible. In Denmark, the Danish Building Regulation (BR10) (Danish Enterprise and Construction Authority, 2010) that was enforced June 30, 2010 by the Danish Ministry of Economic and Business Affairs, stipulates that inhabitable rooms in newly built DCCs must be ventilated by installations comprising of both forced air supply and exhaust. Furthermore, BR10 stipulates that the outdoor air supply rates should not be 1000 ppm, and more than half of the children attending these day cares were exposed to CO2 concentrations exceeding 2000 ppm for a 20min period (Clausen et al., 2009). Finally, the results of that study indicate that high CO2 concentrations, occupant density and children’s sick leave were strongly correlated (Clausen et al., 2009, 2011, 2012). Nursery children were not included in that study, but the reported effects are expected to be stronger among nursery children, as their immune systems are not fully developed (Nafstad et al., 2004). The importance of ventilation has been highlighted in several comprehensive scientific reviews on the effects of ventilation on health (Sundell et al., 2011; Wargocki et al., 2002). These reviews concluded that there was a strong association between ventilation and airway infections as well as increased short-term sick leave 158
among adults working in offices with outdoor exchange rates below 25 l/s per person. These conclusions are further supported by one study showing that the probability of detecting airborne rhinoviruses was positively associated with weekly average CO2 levels in the office (Myatt et al., 2004), as well as another review that concluded there was strong evidence of an association between ventilation and air movements in buildings with the transmission and spread of infectious diseases such as measles, tuberculosis, chickenpox, influenza, smallpox, and SARS (Li et al., 2007). No international or Danish studies have so far investigated whether there is an association between the air change rate and number of sick leave days among children in DCCs due to infectious disease. It is well established that low air exchange rates result in increased concentration of indoor-generated pollutants that are associated with symptoms of ‘sick building’ syndrome and other discomfort and adverse health effects (Sepp€anen et al., 1999; Wargocki et al., 2002) and similar effects applying to bacteria and viruses are possible. The aim of this study was to investigate whether there is an association between ventilation in DCCs and sick leave due to infectious disease among nursery children.
Methods The study design
The study was linked to an already established ‘Health in Day-care Centres’ (SIB) study (SIB, 2014), which aimed to determine the factors responsible for the spread of infectious diseases among small children in Danish nurseries and kindergartens. The SIB project was mainly focused on hygiene, and ventilation was not included. The ventilation measurements were therefore conducted 1 year later. The study design is shown in Figure 1. Results concerning the effects of hygiene on the spread of infectious diseases, including hand hygiene, cleaning methods and products as well as surface types, have previously been reported elsewhere (Ibfelt et al., 2015a). Day care centers
As a part of SIB project (SIB, 2014), a total of 23 DCCs were randomly recruited within the Greater Copenhagen Area, Zealand (17 DCCs), and Nyborg, Funen (6 DCCs) in the fall of 2011. All these DCCs were contacted again 1 year later for agreement to participate in ventilation measurements. One of these was excluded due to missing information on child age, and two did not give participation consent leaving 20 DCCs and 635 attending children included in this study as depicted in Figure 2.
Kolarik et al. Measurements of ventilation and indoor environmental variables
The ventilation measurements were made 1 year after the registration of sick leave, during January 14, 2013 to March 12, 2013 (Figure 1). Three proxies of ventilation were used in the study: (i) the air exchange rate (ACR) measured with the decay method (ii) the ACR measured by perfluorocarbon tracer gas (PFT) method, and (iii) CO2 concentration. The ACR measurements with decay method were conducted in unoccupied rooms either early in the morning, before children were divided into classes or late in the afternoon, after children were collected. All windows and doors were closed tight, after which carbon dioxide was dosed into the room up to a concentration range of 1500–5000 ppm. The decay of the CO2 concentrations was monitored every 30 s for at least 40 min using CARBOCAPÒ CO2 monitors (GMW115; Vaisala, Vantaa, Finland) connected to HOBO U12 -013 data loggers (Onset, Bourne, MA, USA). The windows and doors remained closed during the measurement. A mixing fan was installed to achieve good mixing of the air during the measurement. The 7-day ACR of each DCC was determined using the PFT method (Bergsøe, 1992; Leaderer et al., 1985). The DCCs were treated as single zones and PFT emitters (6–14 per day care), and capillary adsorption tube samplers (7–21 per day care) were deployed. The number of deployed emitters and samplers was dependent on the size of the measured area in each DCC. The amount of tracer adsorbed in the samplers was analyzed using gas chromatography – electron capture detector (GC-ECD), and the ACR was calculated on the basis of the concentration, measured temperature and known emission rates, as well as the volume of the measured part of the building. The PFT measurements were conducted during normal operation of DCCs. Carbon dioxide was measured with CARBOCAPÒ CO2 monitors associated with external input of HOBO data loggers. CARBOCAP CO2 monitors measure CO2 concentrations in the range 0–5000 ppm with an accuracy of (2% of range + 2% of reading) (Vaisala). The mean CO2 concentrations during opening hours (only working days, from 6 am to 5 pm) was calculated and used as proxy of ventilation. The temperature and relative humidity (RH) were measured continuously during 1 week using HOBO U12 -013 data logger. As given by producer, HOBO loggers measures temperature within range 20 to 70°C with accuracy 0.35°C from 0 to 50°C and relative humidity within range 5–95% with accuracy 2.5% (Onset). Similarly to CO2, mean concentrations were calculated for opening hours only. All measurements were carried out in one playroom and one sleeping room in the nursery part of the DCC. As far as possible, sampling was conducted at a height of 1–1.5 m above the floor and not in immediate prox160
imity of windows, heaters, and doors. However, safety of children was of paramount over optimal setup. Furthermore, visual inspections of buildings and interviews with personnel with focus on ventilation type (mechanical, natural, hybrid), type and age of windows and window opening habits were performed. Finally, it was noted whether the smallest children (0–2 years) slept outside in cots/prams during their midday naptime, which is a common practice in Denmark and would reduce indoor exposure time. Sampling and processing of bacteria and virus
The sampling and processing of bacteria and viruses were conducted in winter 2012 as a part of the SIB project (Figure 1), and it have been previously described in detail elsewhere (Ibfelt et al., 2015b). In brief, gastrointestinal viruses were sampled from surfaces at six locations (three in the playroom and three in the changing room/toilet), airway viruses were sampled from surfaces at three locations (all in the playroom) and bacteria were sampled from surfaces at 15 locations (three in the kitchen, seven in a playroom and five in a changing room/toilet). The results of bacteria measurements in the kitchens were excluded from the analysis as ventilation measurements were not conducted here and children do not have regular access to this room. In contrast, changing rooms and toilets are usually directly connected with the playrooms in Danish DCCs and doors are often open; therefore, these measurements were included in the analysis. More information on sampling procedure and analyzed species can be found elsewhere (Ibfelt et al., 2015a). Statistical analysis
We analyzed the association between ventilation and the number of sick days caused by infectious disease among 635 children attending 20 DCCs in Denmark. We used linear mixed model with number of sick days per child as dependent variable and included DCC as a random factor to account for the fact that children attending same institution are correlated. Separate analyses were conducted for three proxies of ventilation as the main explanatory variable: (i) ACR measured with decay method, (ii) ACR measured with PFT method, and (iii) mean CO2 concentration. The effect of each of the three proxies of ventilation on number of sick days was evaluated in several successive steps including variables considered to be associated with sick leave: (i) model 1: crude model with no adjustments and including DCC as a random factor, (ii) model 2: as for model 1 with further adjustment for gender (female vs. male) and age (3 years), and (iii) model 3: as for model 2 and additionally adjusted for sleeping outside (yes, no)
Kolarik et al. Measurements of ventilation and indoor environmental variables
The ventilation measurements were made 1 year after the registration of sick leave, during January 14, 2013 to March 12, 2013 (Figure 1). Three proxies of ventilation were used in the study: (i) the air exchange rate (ACR) measured with the decay method (ii) the ACR measured by perfluorocarbon tracer gas (PFT) method, and (iii) CO2 concentration. The ACR measurements with decay method were conducted in unoccupied rooms either early in the morning, before children were divided into classes or late in the afternoon, after children were collected. All windows and doors were closed tight, after which carbon dioxide was dosed into the room up to a concentration range of 1500–5000 ppm. The decay of the CO2 concentrations was monitored every 30 s for at least 40 min using CARBOCAPÒ CO2 monitors (GMW115; Vaisala, Vantaa, Finland) connected to HOBO U12 -013 data loggers (Onset, Bourne, MA, USA). The windows and doors remained closed during the measurement. A mixing fan was installed to achieve good mixing of the air during the measurement. The 7-day ACR of each DCC was determined using the PFT method (Bergsøe, 1992; Leaderer et al., 1985). The DCCs were treated as single zones and PFT emitters (6–14 per day care), and capillary adsorption tube samplers (7–21 per day care) were deployed. The number of deployed emitters and samplers was dependent on the size of the measured area in each DCC. The amount of tracer adsorbed in the samplers was analyzed using gas chromatography – electron capture detector (GC-ECD), and the ACR was calculated on the basis of the concentration, measured temperature and known emission rates, as well as the volume of the measured part of the building. The PFT measurements were conducted during normal operation of DCCs. Carbon dioxide was measured with CARBOCAPÒ CO2 monitors associated with external input of HOBO data loggers. CARBOCAP CO2 monitors measure CO2 concentrations in the range 0–5000 ppm with an accuracy of (2% of range + 2% of reading) (Vaisala). The mean CO2 concentrations during opening hours (only working days, from 6 am to 5 pm) was calculated and used as proxy of ventilation. The temperature and relative humidity (RH) were measured continuously during 1 week using HOBO U12 -013 data logger. As given by producer, HOBO loggers measures temperature within range 20 to 70°C with accuracy 0.35°C from 0 to 50°C and relative humidity within range 5–95% with accuracy 2.5% (Onset). Similarly to CO2, mean concentrations were calculated for opening hours only. All measurements were carried out in one playroom and one sleeping room in the nursery part of the DCC. As far as possible, sampling was conducted at a height of 1–1.5 m above the floor and not in immediate prox160
imity of windows, heaters, and doors. However, safety of children was of paramount over optimal setup. Furthermore, visual inspections of buildings and interviews with personnel with focus on ventilation type (mechanical, natural, hybrid), type and age of windows and window opening habits were performed. Finally, it was noted whether the smallest children (0–2 years) slept outside in cots/prams during their midday naptime, which is a common practice in Denmark and would reduce indoor exposure time. Sampling and processing of bacteria and virus
The sampling and processing of bacteria and viruses were conducted in winter 2012 as a part of the SIB project (Figure 1), and it have been previously described in detail elsewhere (Ibfelt et al., 2015b). In brief, gastrointestinal viruses were sampled from surfaces at six locations (three in the playroom and three in the changing room/toilet), airway viruses were sampled from surfaces at three locations (all in the playroom) and bacteria were sampled from surfaces at 15 locations (three in the kitchen, seven in a playroom and five in a changing room/toilet). The results of bacteria measurements in the kitchens were excluded from the analysis as ventilation measurements were not conducted here and children do not have regular access to this room. In contrast, changing rooms and toilets are usually directly connected with the playrooms in Danish DCCs and doors are often open; therefore, these measurements were included in the analysis. More information on sampling procedure and analyzed species can be found elsewhere (Ibfelt et al., 2015a). Statistical analysis
We analyzed the association between ventilation and the number of sick days caused by infectious disease among 635 children attending 20 DCCs in Denmark. We used linear mixed model with number of sick days per child as dependent variable and included DCC as a random factor to account for the fact that children attending same institution are correlated. Separate analyses were conducted for three proxies of ventilation as the main explanatory variable: (i) ACR measured with decay method, (ii) ACR measured with PFT method, and (iii) mean CO2 concentration. The effect of each of the three proxies of ventilation on number of sick days was evaluated in several successive steps including variables considered to be associated with sick leave: (i) model 1: crude model with no adjustments and including DCC as a random factor, (ii) model 2: as for model 1 with further adjustment for gender (female vs. male) and age (3 years), and (iii) model 3: as for model 2 and additionally adjusted for sleeping outside (yes, no)
Ventilation and sick leave in day care centers and municipality (Copenhagen area vs. Nyborg) (Figure 3). Additionally, several sensitivity analyses (models 4– 6) were conducted. Model 4 was fit as model 3 and additionally adjusted for variables which may directly affect three ventilation measures: frequency of windows opening (1, 2, and 3 or more times per day), age of windows (newer with double glazing vs. older with single glazing and/or inner sash) and type of windows (wooden vs. plastic). Model 5 was fit as model 4 and additionally adjusted for mean temperature and RH, which are related both to ventilation measures as well as to sick leave. Finally, model 6 was fit as model 4 and additionally adjusted for possible intermediate variables or proxies of sick leave: gastrointestinal or airway virus and nasopharyngeal or intestinal bacteria (Figure 3). We did not have decay measurements in DCC 2 and 8 (Table 2); thus, the analyses of the effect of decay were based on data excluding these DCCs. We repeated therefore the analyses based on a subset of data excluding DCC 2 and 8, in which we explored the effect of the proxies CO2 and PFT. The significance threshold was P < 0.05, and all analyses were performed in SPSS (IBM SPSS Statistics 22 for Windows, Armonk, NY, USA). Within two of the included DCCs, nursery children, and kindergarten children shared the same rooms (in so called family rooms), and for these two DCCs, the older children (above 3 years) were also included in the analyses. The sick leave data showed skewed distribution; therefore, all regression analyses were performed using the natural logarithm of the number of sick days. The association between sick days caused by infectious disease and each of the ventilation proxies was calculated
Crude model
Model 1: Crude model Day-care as a random factor
Main models
Model 2: Fitted as the crude model with adjustment for gender and age
Sensitivity analyses
Model 4 Fitted as Model 3 with further adjustment for frequency of opening windows, age of windows and window type
by exponential transformation of the regression estimate and expressed as a percentage change. Results
This study included 482 nursery children aged 0– 3 years and 153 kindergarten children aged over 3 years. The included DCCs had 1–5 nurseries (or family groups), with 11–34 children in each group (median 12). The total number of groups in each day care (including both nursery and kindergarten groups) ranged from 2 to 7, and the number of children per DCC ranged from 24 to 149. All but two DCCs had balanced mechanical ventilation systems. The description of the study population and the investigated DCCs is presented in Table 1. Results of temperature, RH and CO2 measurements, as well as ventilation measurements for each DCC are presented in Table 2. The median temperature and RH during measuring period were 22°C and 27%, respectively, and the median concentration of CO2 calculated for opening hours only was 579 ppm (mean 643 ppm). The maximum 20 min running average during opening hours was 1132 ppm, ranging from 681 to 2864 ppm. In 9 DCCs, children experienced at least one 20-min period with CO2 concentrations exceeding 1000 ppm and in one DCC with concentration over 2000 ppm (data not shown). The ACR differed between the DCCs (Table 2) and between measurement techniques. There was no correlation between air change rate measured with PFT and decay method (Pearson coefficient 0.082, P = 0.74) nor was PFT significantly correlated with mean CO2 concentration (Pearson’s coefficient 0.233, P = 0.31). However, there was a significant negative correlation between ACR measured with decay method and mean
Model 3 Fitted as Model 2 with further adjustment for sleeping outside and municipality
Model 5 Fitted as Model 4 with further adjustment for mean temperature and relative humidity which are both related to ventilation measures and sick leave
Model 6 Fitted as Model 4 with further adjustment for possible intermediate variables or proxies of sick leave
Fig. 3 The statistical approach
161
Kolarik et al. Table 1 Characteristic of the study population and the investigated DCCs
Characteristic Room type at DCCa Nursery (children 0–3 years) Family rooms (children 0–6 years) Number of children per day care Child gendera Male Female Child age (years)a 3 Sick leave (days)b Sleeping outsidec Yes In crib rooms No Window agec New Old Window constructionc Wooden Plastic Window opening frequencyc Once a day Twice a day ≥Three times a day Ventilation typec Mechanical Natural Municipalityc Great Copenhagen area, Zealand Nyborg, Funen Samples with detected airway or intestinal bacteriad Samples with detected airway or gastro-viruse
N (%)
Median (5th, 95th percentile)
482 (75.9) 153 (24.1) 36 (12, 90) 331 (52.1) 304 (47.9) 0 (0) 220 (34.6) 273 (43.0) 142 (22.4) 3 (0, 13) 4 (20) 13 (65) 3 (15) 15 (75) 5 (25) 11 (55) 9 (45) 9 (45) 9 (45) 2 (10) 18 (90) 2 (10) 15 (75) 5 (25)
Discussion 4 (0, 6) 16 (8, 20)
a
Data per child (635 children included in analyses). Based on total number of sick-leave days of all included children during 11-week period. c Data per DCC (20 day cares included in analyses). d Maximum 24, see Methods for specification. e Maximum 78, see Methods for specification. b
CO2 concentration (Pearson coefficient 0.691, P = 0.001), such that lower CO2 concentrations were measured in DCCs with higher ACR. The highest number of sick days was recorded in DCCs with lowest ACR measured with the decay method, as well as among boys as compared to girls and among smallest children as compared to older children (results not shown). The inverse relationship between the number of sick days recorded in winter 2012 and ACR measured with decay method in winter 2013 was statistically significant in the crude analysis and the two main adjusted models (model 2 and 3; Table 3), with a 12% (95% confidence interval 2–21%) decrease in the number of sick days per hour increase in ACR. When the variables affecting ventilation rates in a room (model 4) and the intermediate variables (model 6) were included into the model, the association 162
was only slightly attenuated, losing however statistical significance, while adjustment for temperature and RH (model 5) resulted in an even stronger and more statistically significant associations (Table 3). The associations between sick leave and the other two proxies of ventilation measured 1 year after sick leave registration were not statistically significant. The results for CO2 concentration measurements point however to the same direction, showing a 2% increase in sick leave (not significant) per 100 ppm increase in CO2 concentration, while results for ACR measurements with PFT are opposite in sign, such that more sick days have been registered in DCCs with higher ACR measured with PFT (not significant; Table 3). In subanalyses excluding DCCs 2 and 8, which were missing in our analyses of decay, we did not see an effect of CO2 or PFT. The estimates were as follows: PFT: 10.9% change with CI: 4.4 and 28.7 and CO2: 1.8% change with CI: 3.3 and 7.1; P = 0.512 (results not shown). None of the proxies of ventilation were significantly correlated with number of bacteria or virus detected on indoor surfaces 1 year before ventilation measurements (results not shown). A non-significant negative relationship was found between ACR measured with decay and gastro-virus as well as airway bacteria and airway virus, while a non-significant positive relationship was found between ACR measured with decay and intestinal bacteria (results not shown).
This study linked data on sick leave among 635 children attending 20 DCCs with ACR and CO2 measurements from these DCCs conducted a year later. Mean concentrations of CO2 were relatively low in this study, and 1000 ppm was only exceeded in one DCC. Air change rates varied among DCCs and between measuring techniques. Results from this study show an association between sick leave among nursery children and ventilation in DCCs measured with decay method. To our knowledge, the only other previous study investigating sick leave was also conducted in Denmark (Clausen et al., 2011) and reported a significant decrease in sick leave among kindergarten children (aged 3–6 years) associated with a decrease in the maximum 20 min average CO2 concentration (Clausen et al., 2011). Although results are not directly comparable with ours as we studied mainly nursery children (0–3 years of age), we report the same trends in the association between sick leave and CO2, showing a 2% increase (not significant) in sick leave per 100 ppm increase in CO2 concentration. In addition, previous results reported in studies in school and office environments corroborate this association (Ervasti et al., 2012; Milton et al., 2000; Shendell et al., 2004; Simons et al., 2010).
Ventilation and sick leave in day care centers Table 2 Temperature, relative humidity, concentration of carbon dioxide and air change rate (ACR) during measuring period
Day care 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Reference
Ventilation (N-natural, M-mechanical)
Temperature, °C Median (5th, 95th percentile)
M M M M M M M M M M N M M M M M M M N M
22.5 (21.2, 24.5) 21.5 (20.2, 22.3) 21.3 (20.3, 22.1) 22.6 (20.8, 24.3) 24.1 (23.5, 25.9) 21.7 (20.9, 22.5) 21.0 (20.1, 22.0) 25.7 (20.9, 27.2) 21.7 (19.8, 23.1) 22.0 (21.0, 22.9) 20.3 (17.9, 21.6) 21.4 (20.7, 22.1) 19.8 (19.1, 20.4) 21.8 (20.8, 22.8) 21.7 (19.1, 22.9) 22.5 (20.6, 24.4) 23.9 (23.2, 24.2) 21.9 (21.2, 22.8) 20.2 (19.6, 21.1) 22.5 (21.4, 23.3) Summer 23°C
