Clustering of office workers from the OFFICAIR study in The Netherlands based on their self-reported health and comfort

The growing field of indoor health and comfort studies recently shifted from predicting the average comfort and wellbeing of a large population into identifying the needs of individuals in different scenarios. This study aimed to identify different profiles of office workers in the Netherlands who took part in the OFFICAIR study, based on their self-reported health and comfort. Associations of respondents’ health and comfort with gender and type of office indicated that female occupants experienced significantly higher numbers of building-related symptoms and consistently lower satisfaction levels of their office environment than male occupants. Workers in open space offices without partitions reported lower satisfaction and suffered from building-related symptoms more frequently than occupants in single person offices. TwoStep cluster analysis revealed three profiles of occupants: Healthy and satisfied workers, Moderate healthy and noise-bothered workers and Unhealthy and Air and temperaturebothered workers. While the first group was by far the healthiest, significant higher risks for building-related symptoms such as dry eyes (OR: 3.38), dry skin (OR: 2.87) and watering, itchy eyes (OR: 2.7) were identified for the unhealthy group than for the moderate healthy group. The results confirm the need of an integrated approach to better understand moderate and unhealthy groups in order to provide customised solutions for individuals with different complaints and needs.


Introduction
Office workers spend a large part of their time inside the buildings in which they work, therefore office buildings should ensure comfort, health and wellbeing for their occupants'. In fact, 'a healthy, energyefficient and comfortable indoor environment' has been researched by many studies for over several decades [1]. Previous studies clearly suggest a direct link between building design (including indoor building conditions) and human health and well-being of the occupants [1][2][3][4][5][6]. For example, an increase in thermal satisfaction raised comfort expectations of other indoor environmental quality (IEQ) factors, and had a positive effect on occupants' productivity in a study performed by Geng et al. [7]. In several studies, employees with higher lighting appraisals reported a better mood and improved health and well-being at the end of their workday [8,9]. It has been shown that lighting quality plays an important role in the control of the day-night rhythm [10,11]. Several indoor air quality (IAQ) studies highlighted various health-related issues reported by some of the building occupants who were dissatisfied with IAQ [3,12,13]. Unfortunately, many of these findings have not yet been reflected in present-day IEQ guidelines to prevent or reduce health and comfort effects of occupants.
With respect to the relative impact (or importance) of different indoor environmental conditions on office workers' level of overall satisfaction with the office environment, mixed outcomes are often found [14,15]. The most important (or influential) environmental aspect can differ per study: the amount of space and acoustics in Ref. [16], air quality in Refs. [17,18], thermal environment in Refs. [19][20][21][22][23], privacy and acoustics in Refs. [24], and acoustics and office layout in Ref. [25]. There are many differences between the studies listed above such as the study design (e.g. building type, location and construction date) [2,26] and social-cultural factors [19,27], that may account for these inconsistent findings. It is also suggested that people experience and complain about certain building-related symptoms even when their workspaces comply with the guidelines [2,28].
The lack of consistency in some findings are fundamentally due to the fact that the built environment and its indoor environment with occupants is a complex system with many interrelations, that can be linear or non-linear [29,30]. Yet, IEQ is still often assessed mainly on single dose-related indicators, and is developed for the average occupant, ignoring the fact that we are dealing with individuals in different scenarios and situations. To overcome the current difficulties, recently, an integrated analysis approach was introduced, which takes account of the combined effects of various (positive and negative) stressors and includes occupants' preferences and needs as well as interactions at human and environmental level [28]. The goal of the integrated analysis approach is to match different profiles of people with different patterns of positive and negative stressors for a certain situation, instead of linking indoor environmental conditions to average comfort and health of a large population. To achieve this, profiling of occupants based on their comfort, behaviours and needs of IEQ in certain scenarios and situations is essential.
Several studies have clustered occupants based on their behaviours and preferences with different methods. For example, the K-mean clustering algorithm was used to profile occupants based on their lighting control behaviours and dimming level preferences [31]; and it was used to profile office workers based on electricity load patterns [32], HVAC-schedules [33] and thermal preferences [34]. The TwoStep clustering algorithm has been applied particularly in recent studies. Six clusters of primary school children in the Netherlands based on their preferences and needs of IEQ in their classrooms were identified [35]; and five different archetypes of home occupants in the Netherlands based on their comfort behaviours and energy use pattern were identified [36,37]. All of the studies listed above clearly indicate that people can indeed differ in their preferences and needs and that it seems possible to distribute them into clusters (profiling).
Therefore, this study started with questioning whether profiles of office workers can also be identified based on their self-reported health and comfort regarding their office environments. If so, what would be the characteristics of the profiles? In order to answer these questions, this study used existing survey data (from the OFFICAIR study) [2], with a particular focus on the office workers studied in the Netherlands. Because self-reported health and comfort of the respondents in the office buildings in the Netherlands from the OFFICAIR study have not been reported yet, the health and comfort data and their possible associations with some demographic profiles such as gender and type of office, were explored first. Then, clustering was performed.

Study population and design
OFFICAIR was a European research project, which involved collecting data from 167 office buildings in eight European countries (Greece, France, Finland, Hungary, Italy, Portugal, Spain and the Netherlands) during the winter of 2011-2012, and questionnaire data from 7441 office workers. The procedure of this cross-sectional study is described in detail elsewhere [2], and is therefore only briefly summarized here. This study focuses on the results from the Netherlands, in which in total 1014 office workers from 20 different office buildings participated. All the buildings shared the following characteristics: (1) new or recently retrofitted buildings (less than 10 years old) at the start of the study, (2) buildings had been operating in their current form for a minimum of 1 year prior to the start of the study, (3) no major renovation was planned within a year at the start of the study.

Worker questionnaire
Self-reported data, collected by questionnaires, were used to investigate office workers' health and comfort. Health symptoms and experienced comfort of the respondents in this study are explained and summarized in Table 1 and Table 2, respectively.
Concerning health symptoms (a total of 20), participants were asked: "During the past four weeks, on how many days did you experience each of the following symptoms when you were at work at your workstation (including today)?" These questions were answered by occupants using a five-point scale. Health symptoms that office workers frequently experienced during the past month were the primary focus of this study. Frequent occurrence was defined as 'greater than or equal to 1-3 days per week in the last 4 weeks'. Then the item was dichotomized into: "Did you experience the following symptoms frequently during the past four weeks?" (Yes/No) (see Table 1). In the result section, both values ('frequently experienced' and 'experienced') are presented (see Fig. 1) which allows comparison of the results with the complete dataset of the OFFICAIR study [2,38].
For comfort with environmental conditions (in total 17), participants were asked: "How would you describe the typical indoor conditions in your office environment during the past four weeks?" These questions were originally answered by the occupants using a seven-point scale. In previous publications on the OFFICAIR results [2,38] the first three values (of the seven-point scale) were used to define dissatisfaction, whereas in another publication the first two values were used to define clear dissatisfaction [39]. In this study the first two values were converted into the percentages of office workers who expressed clear dissatisfaction with the indoor environmental conditions (see Table 2).

Participants
3569 invitation e-mails were sent, 1319 respondents started and 1022 submitted the questionnaire. A total of 1014 completed questionnaires were found to be valid and therefore were included in this study. Table 3 shows some characteristics of the respondents. Workers were balanced in terms of gender (ratio male/female: 1.02). The mean (standard deviation) age of the respondent was 43.8 (10.1) years. Most of workers (69%) had a university degree. Concerning their life style,

Data management and analysis
The percentage of occupants who expressed clear dissatisfaction was calculated for each of the 19 variables (converting the dichotomized scale increased the number of conditions from 17 to 19). Percentage of dissatisfied (defined as the cumulative total of the first two points on the seven-point satisfaction scale) is regarded as a meaningful and practical metric in IEQ research [3] and has been adopted in several other studies and therefore is directly comparable to that used by other researchers [3,40,41]. A series of multivariate logistic regression analyses were performed to investigate the association between dissatisfaction with the different environmental conditions and personal and work-related factors. The same approach was used to investigate the relationship between the 20 self-reported symptoms and personal and work-related factors. The results of the logistic regression models are reported as adjusted odds ratio (AOR). The AOR represents the ratio of the odds of a particular outcome occurring in one group over the odds of it occurring in another group.
The TwoStep cluster analysis approach was then used to categorize the office workers based on their self-reported health and comfort. The TwoStep cluster analysis was selected for the following reasons. First, it permits both continuous and categorical data to be analysed simultaneously. Second, optimal number of clusters are automatically selected by the method. Thirdly, the method is suitable for large data sets [42,43]. Several recent studies also showed that the TwoStep cluster analysis was an effective tool to cluster school children [35] and home occupants [36,37]. For the detailed setting of the TwoStep analysis, the option of optimum number of clusters, log-likelihood distance measure and Akaike's Information Criterion were selected. As suggested by Noru� sis [43], the following steps were performed to validate the final solution model. First, the silhouette measure of cohesion of the clusters were checked (value recommended to be above 0.0 and preferably 0.2). Second, Chi-square tests and t-tests were performed with categorical and continuous variables respectively: all variables in the solution needed to be statistically significant (p < 0.05). Third, predictor importance scores were checked (recommended to be greater than 0.02). Finally, a series of multivariate logistic regression analyses were performed again to compare the strength of the associations between different identified clusters and occupants' health and comfort quality. All of the statistical analysis used in this study were conducted using IBM SPSS 24.

Health symptoms
In Fig. 1 the percentages of office workers having symptoms are presented. Three of the most prevalent symptoms for 'frequently experienced', (and 'experienced') were: 'dry eyes' (31% 'frequently experienced' and 47% 'experienced'), 'dry skin' (23% and 30%) and 'burning, irritated eyes' (18% and 29%). More than one-seventh of the workers in the Netherlands reported that they had 'frequently experienced' (and 'experienced') 'blocked nose' (17% and 30%), 'headache' (16% and 40%), 'dry, irritated throat' (16% and 27%) and 'sneezing' (14% and 24%) at their workstations during the last month. While 64% and 48% blamed no particular season or no particular of day, more than one in five of the workers (23% and 42%) answered that symptoms tended to be worst in the winter season and in the afternoon of a day, respectively. 76% of the workers believed that any of the experienced symptoms was due to their office environment.
Multivariate logistic regression analysis was conducted to investigate the associations between the 10 highest prevalent symptoms and confounding factors gender and type of office. Table 4 presents the adjusted odds ratios (OR) for each of the 10 symptoms. The outcome showed that female workers in the Netherlands had significantly more symptoms than male workers. 'Dry skin' showed the strongest association (36% vs. 10%; OR ¼ 4.19, 95% CI ¼ 2.89-6.08, p < 0.001) with female gender, followed by 'headache' (24% vs. 8%; OR ¼ 3.00, 95% CI ¼ Respondents occupying an 'open space without partitions' were significantly associated with higher rates of four symptoms (headache, dry/irritated throat, dry eyes, dry skin) than respondents occupying a 'single person office' (Table 4). More specifically, the result showed that subjects who worked in an open space without partitions were 2.25 times more likely to experience 'headache' (24% vs. 9%' OR ¼ 2.25, 95% CI ¼ 1.23-4.11, p ¼ 0.008) and 'dry, irritated throat' (22% vs. 10%' Table 2 Self-reported dissatisfaction with environmental conditions and the scales used.
Different types of offices were also found to be associated with perceived environmental conditions. Open space without partitions was significantly associated with higher rates of clear dissatisfaction on many of variables than in single person offices. Two variables that particularly showed the strongest relationship with 'open space without partition' were 'overall thermal comfort' (40% vs. 9%; OR ¼ 5.43, 95% CI ¼ 3.05-9.66, p < 0.001) and 'draughty air movement' (28% vs. 7%; OR ¼ 5.39, 95% CI ¼ 2.83-10.26, p < 0.001). Type of office was not found to be correlated with clear dissatisfaction with 'natural light' or 'artificial light'. Office workers in open spaces (either with partition or without partition) were more likely to express clear dissatisfaction with 'reflection or glare' than office workers in single person offices.

Clustering of office workers
The TwoStep cluster analysis resulted in three clusters with 909 workers (105 workers, as incomplete sample, were automatically excluded by the two-step cluster analysis). Initially, 18 variables (excluding 'overall comfort') were included and two variables (air--humid and noise-outside) were removed when a final cluster solution was achieved, as their prediction score was lower than 0.02 [10]. The final solution presents a silhouette measure of cohesion and separation of 0.3, which ensures that within and between-cluster distance is valid amongst the 16 variables, indicating variation between the variables. Comparison of means analysis ensured that the final 16 variables were statistically significant, and hence they varied between clusters. Additionally, the variables with the lowest score for predictor importance was found to have a rating of 0.07, above the recommended 0.02. The predict importance of these variables in the final solution were: air-dry (1.00), noise-inside (0.80), air-overall quality (0.60), noise-overall quality (0.37), light-overall quality (0.25), light-artificial (0.20), light-natural (0.13), air-stuffy (0.11), temperature-too much variation (0.10), temperature-too cold (0.07), light-glare (0.07), noise-building system (0.07), air-smelly (0.07), air-draughty (0.07). All of these variables were confirmed to be statistically significant related to the three clusters.  Table 6 presents general characteristics of the office workers such as age, gender, type of office, type of work, perceived health and comfort for the three identified clusters. Mean age of the workers in each cluster showed no difference, while significant statistical difference in gender, type of office and type of work were noticed. In terms of perceived health and comfort, office workers from three clusters varied significantly. All of the variables in these categories showed statistically significant differences. To investigate the differences in their self-reported health and comfort more thoroughly, a series of multivariate regression analyses with controlling age, gender, type of office and type of work, were performed. Tables 7 and 8 present the adjusted odd ratios (OR) with 95% confidence intervals for association of three clusters with   ¼ 909), and comprises of a high proportion of men (68%), more or less equally distributed among the different age groups. Most of the office workers are highly educated: 44% holds a degree that is equal or higher than a master degree. A significant number (65%) used  P-values are from Bonferroni adjustments made after chi-square tests.

Table 7
Logistic regression analysis for association of symptoms for the three clusters of office workers. their own car for commuting to work. 50% of them reported to have at least one disease or problem. The most reported disease was 'allergy' (23%), followed by 'eczema' (9%) and 'high blood pressure' (9%).

IEQ-related complaints.
Except for the fact that 28% of them described that their offices were 'too dry', generally very few (less than 10%) expressed clear dissatisfaction with the IEQ of their office. Almost none (0.3%) reported strong dissatisfaction with the overall comfort. As shown in Table 6, cluster 1 reported statistically significant less number of IEQ-related complaints compared to the other two clusters. The only exception was 'humid air', for which almost no one reported a complaint. Overall, the results suggest that the office workers grouped as cluster 1 were not bothered by the IEQ of their office environment.
3.3.1.1.3. Health symptoms. The most prevalent building-related symptom observed in this group was 'dry eyes' (12%). 10% frequently experienced 'sneezing' and 9% reported complaints about 'dry skin' and 'blocked, stuffy nose'. Except for 'sneezing' (no statistically significant difference between cluster 1 and cluster 2 was found), cluster 1 experienced building-related symptoms in the last 4 weeks significantly lower than the other profiles of the workers even after controlling for age, gender, type of office and type of work (see Table 7).

Work-related characteristics.
The majority (75%) of cluster 1 worked in a (either single or shared) office, 50% of them answered that they shared their office with only one person (see Appendix A), and most reported type of work was 'professional' (33%). Compared to the average value of all the clusters, more 'managerial' (27% vs. 21%) and less 'clerical-secretarial' (20% vs. 25%) were observed. Almost all (97%) reported to work with a visual display unit (VDU) and average hours per week with a VDU at work was 25 h. Similar results were found in the other two clusters. Cluster 1 had the lowest mean ERI-value, which suggests that their work-related stress is perceived lower than in the other clusters.
3.3.1.1.5. Building and office characteristics. Cluster 1 respondents were most frequently located in small-sized office buildings (typical number of occupants <170: 70%; total floor area <1500 m 2 : 43%), located in either industrial area or suburban or village in a rural area (51%), and most likely, therefore, exposed to relatively fewer sources of outdoor noise or air pollution than the office workers in the other clusters. The majority (80%) worked in offices with operable windows, carpet flooring (87%), mineral fibre ceiling tiles (84%) and dispersion, emulation paint on the wall (81%). Floors in their offices were cleaned at least once per week (72%) and almost no (6%) visible mould growth in their offices was spotted.

Cluster 2: Moderate healthy and noise-bothered workers
3.3.1.2.1. Personal characteristics. Cluster 2 represents 33% of the total sample and is balanced in terms of gender ratio (52% women and 47% men). This cluster tends to be younger (40% younger than 40) than the others and irrelevant to their educational background. Allergy (17%), eczema (12%) and high blood pressure (10%) were three most reported diseases. Car (55%) and bike (50%) were chosen most often for commuting to work.

IEQ-related complaints.
This cluster had the highest percentage of noise complaints. 37% (vs. 18% total sample) were clearly dissatisfied with the overall noise quality and 63% (31% total sample) reported noise from inside (e.g. colleagues chatting or phone calls). A significant number of them also complained about IAQ. 34% were clearly dissatisfied with overall air quality, 53% with dry, 28% with stuffy and 25% with draughty air. Almost one-third complained about 'overall thermal comfort', 'too much variation of temperature' as well as 'too cold'. Light quality was least of their concerns: one-in-five were dissatisfied with their lit environment.

Health symptoms.
31% of this group experienced dry eyes symptoms more than once per week in the last month. Headache (21%), dry skin (21%) and burning, irritated eyes (19%) were reported. As can be seen in Table 7, except for 'sneezing' (15%), this group experienced a significantly higher number of health-related symptoms than Cluster 1. Interestingly, their average reported number of health-related symptoms are very close to the averages reported by the total sample (see Table 6).

Work-related characteristics.
48% of the respondents in Cluster 2 worked in an open space without partition; the number of occupants in the working space varied between 2 and 20 (median: 7). Compared to the average value of the total sample, a higher number of 'professionals' (39% vs. 32%) was seen compared to the total sample. Their average ERI score (and percentage of people who scored above an ERI of 1) was the highest among the three clusters, which suggests that this group seems to be under relatively high work-related stress.
3.3.1.2.5. Building and office characteristics. Cluster 2 respondents most frequently worked in large-sized buildings (typical number of occupants >170: 61%; total floor area >2500 m 2 : 59%), located in the city centre (47%). For 97% of the buildings, sources of nearby outdoor noise and 99% outdoor air pollution were observed. A significant number of the respondents (39%) worked in buildings in which windows were not operable. Office covering and furnishing were found to significantly differ from the office environment of Cluster 1. A relatively higher number of them worked in offices which had synthetic smooth floor covering (27%), ceiling tiles (22%) and exposed concrete or plastered wall (17%). Although surfaces were cleaned at least once per week in 68% of the offices, in several of them (22%) visible mould growth was observed.

Cluster 3: Unhealthy and air and temperature-bothered workers
3.3.1.3.1. Personal characteristics. This cluster comprises of a high proportion of women (74%), while age is equally distributed over the three categories. With regards to their educational background, this group is different from the other two clusters; they consist of more 'professionals' and less workers with a 'master, PhD or specialization'. Also, they suffered from more diseases than the other clusters: workers suffering from 'allergy' (30%) and 'migraine' (13%) was found to be significantly higher. Apart from the above factors, workers in this cluster share very similar characteristics as the workers in Cluster 2.

IEQ-related complaints.
This group in general complained a lot about IEQ. 63% of them were clearly dissatisfied with the 'overall Note. Adjusted odds ratio implies controlling for gender, age, type of office and type of work. ***p < 0.001, **p < 0.01, *p < 0.05, NS ¼ not significant (p > 0.05).
air quality' in their offices, 100% found the air 'too dry', 46% 'too stuffy', 27% 'too draughty' and 20% 'too smelly'. But also with the thermal conditions they were dissatisfied: 48% expressed clear dissatisfaction with overall thermal quality and 30% with 'temperature too variable', 25% found it 'too cold' and 26% 'too hot'. Acoustical quality was also a problem to them but not as much as Cluster 2 did. 39% reported 'too much noise from inside of their offices' and 25% were clearly dissatisfied with the 'overall noise quality'. The results presented in Table 8 show that this group was statistically significant more bothered by air-related factors and thermal-related factors, but less bothered by noise-related factors as compared to Cluster 2.

Health symptoms.
In terms of health-related symptoms, this group can be described as the unhealthiest one. 61% and 48% experienced 'dry eyes' and 'dry skin' at least once per week in the last month, respectively. Multivariate logistic regression shows ( Table 7) that -except for 'headache', 'sneezing', 'watering, itchy eyes' and 'lethargy, unusual tiredness'-this group reported the highest percentages of symptoms.

Work-related characteristics.
Compared to the average values of the total sample, this cluster tends to work more in an open space without partition (50% vs. 36%) and less in a private office (38% vs. 54%). The number of occupants they are sharing their workspace with, varied from 3 to 25 (median: 8). There is a tendency that this group comprised more 'clerical-secretarial' (37% vs 22%) and less 'professional' (23% vs. 32%) workers, compared to the total sample.
3.3.1.3.5. Building and office characteristics. Cluster 3 workers are most frequently observed in large-sized buildings (typical number of occupants >170: 61%; total floor area >2500 m 2 : 63%) located in the city centre (49%). Sources of outdoor noise (96%) and air pollution (99%) existed nearby the buildings. A significant number of workers in this cluster (44%) worked in buildings in which the windows were not operable. Office covering and furnishing were found to be significantly different from the typical office environment of Cluster 1 workers. A relatively higher number of them worked in offices with a synthetic smooth floor covering (18%), ceiling tiles (33%) and exposed concrete/ plastered wall (11%). In 56% of the offices, the surfaces were cleaned at least once per week (56%) and in several of them (33%) visible mould growth was observed.

Comparison to the European-wide OFFICAIR study
In this study, a clear gender effect on self-reported comfort and health from the office workers studied in the Netherlands was found. Compared to the European-wide OFFICAIR study [2,25,38] (see Fig. 3), the female workers in the Netherlands were generally more dissatisfied with their indoor environmental qualities, except for overall noise quality. For self-reported comfort, a gender effect was observed for both the European-wide OFFICAIR study and the OFFICAIR-study performed in the Netherlands. Unlike female workers, male workers in the Netherlands showed very similar patterns to the European-wide male workers.
Regarding building-related symptoms such as dry eyes and headaches, also a gender effect was observed in both the European-wide and the study in the Netherlands. Also, higher percentages of female workers in the Netherlands recorded complaints on dry eyes and headaches than female workers at European-wide level. For male workers, dry eyes were more reported in the Netherlands, while reported percentages of headache seemed almost the same.
It was also seen that open-plan offices were associated with higher numbers of comfort and health-related complaints than private offices. To check whether a consistent result was reported from the OFFICAIR (European-wide) study, a comparison of the results is made in Fig. 4. The results from both studies first suggest that open-plan offices were also associated with higher numbers of comfort and health-related complaints than private offices. But overall, open-plan offices in the Netherlands had much higher percentages of comfort and health-related complaints than open-plan offices from the European-wide study. Additionally, a significantly higher percentage of dry eyes complaints was reported by the occupants in private offices in the Netherlands, while there was almost no difference in percentages of dissatisfied workers with respect to their IEQ-factors as compared to the Europeanwide OFFICAIR study.

Comparison to other office building studies
The results of this study also confirm several findings from previous studies. De Dear et al. [39] observed significant gender differences in office environments for almost all IEQ factors (including thermal, air, lighting, acoustical quality, office layout, furnishings and cleanliness and maintenance), but not for the overall rating of their workspace environment. In this study, not only were females found to be consistently less satisfied with the indoor environmental conditions, including the overall comfort level of the offices, but females were also associated with a significantly higher number of most of the symptoms, particularly 'dry eyes', 'dry skin' and 'headache'. The result of this study shows that gender difference was particularly significant for overall thermal comfort, which confirms findings of previously published field research [39,44,45]. However, these gender effects can also be attributed to other factors such as differences in clothing and metabolic rate, which are often claimed to have an impact on the differences [45,46].
Further, the effect of office layout on self-reported health and comfort, which was found in this study, is in line with findings of several previous studies [47][48][49]. In this study, it was observed that office workers in open-plan layouts reported significantly higher numbers of 'headache', 'dry, irritated throat', 'dry eyes', and 'dry skin' symptoms as well as the fact that private single person offices clearly outperformed open-plan layouts with respect to most IEQ-aspects. As Pejtersen et al. [49] summarized, these differences may be due to a) higher exposure to noise in open-plan office, b) differences in ventilation systems, c) exposure to viruses which presumably is higher in open-plan offices, d) differences in the psychosocial working environment, and e) presence of other humans when working which might lead to lower employee autonomy.

Strengths weaknesses and recommendations
This study is a first attempt to identify different profiles of office workers based on similarities in their IEQ-related complaints through the analysis of an existing dataset of the Dutch offices investigated in the OFFICAIR study. As a result, detailed descriptions of the profiles which include personal characteristics, life style, health symptoms, work characteristics and building/office characteristics were provided. The results have shown particularly large differences between profiles with regard to their health symptoms at work.
To validate the findings better, further studies are recommended as the current study includes a few limitations. First, since the profiles of office workers in this study were created mainly through analysis of selfreported complaints, it would be beneficial to include additional information such as user preference, needs and control behaviours for further studies when designing their data gathering stage. In particular, as several recent studies [35][36][37] have shown that users' needs and preferences were key determining items for their final models of school children and home occupant archetypes, it is expected that such data also is important for enhancing our understanding of the profile of office workers when included.
Additionally, based on the findings of this study, a few recommendations can be made to architects, facility managers and researchers who are keen on designing (or achieving) a healthy office environment. The study suggests different individual specific and context-relevant priorities for different profiles of office occupants rather than devising a fit-for-all solution. For example, providing acoustical privacy panels to the moderately healthy and bothered by noise group is likely to result in a decrease in IEQ-related complains, whereas the same approach might not be so effective to other profiles. Secondly, there is a need for further human-oriented research to better understand the different office workers. In particular the moderate healthy and noise-bothered workers and unhealthy and air and temperature-bothered workers for whom their work characteristics and office/building characteristics didn't differ so much. For example, quantifying and assessment of exposure to different air quality, acoustical and lighting condition for longer period at a personal level would be desirable.

Conclusion
In conclusion, this study raised a particular concern for female workers and open-plan offices in the Netherlands, as their self-reported comfort and health were in general much worse than the average scores from the European-wide OFFICAIR study. Then, three different profiles of office workers in the Netherlands based on their self-reported health and comfort were identified using a large database that consisted of both office workers' comfort and health reports and characteristics of their office buildings. The results indicate that office workers can be grouped into the ones who are satisfied with their indoor environments, the ones who complain about indoor noise, and the ones who are bothered by indoor air and temperature. While the satisfied workers were by far the healthiest among the groups, significantly higher health risks were identified for the office workers who complained about indoor air and temperature than the ones who were bothered most by indoor noise. As the outcome confirmed, there is a need of an integrated approach to better understand the different office workers, in particular the moderate healthy and noise-bothered workers and unhealthy and air and temperature-bothered workers in order to be able to provide customized solutions for their complaints.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgement
This study was sponsored by the fellowship of Philomena Bluyssen provided by the Delft University of Technology under the chair of Indoor Environment. Fig. 4. The percentage of dissatisfied occupants at private office/open-plan office (subjects who rated their satisfaction level with the lowest 2 levels on the 7-point scale) for each IEQ factors and the percentage of the occupants who experienced dry eyes or headache at least once per a week in the past four weeks between the OFFICAIR (European-wide) and the OFFICAIR (The Netherlands).