• No results found

A number of analyses considered factors that may have influenced the experimental outcomes and the ANOV A revealed some differences between these data sets. These are discussed below.

1. Screening non work related symptoms

The analyses undertaken compared the complete and screened data sets to determine if symptoms differed with respect to lighting treatments and offices. In eight of the nine analyses, the F test values were lower for lighting treatment differences in the screened data sets. Mean values remained comparable or were slightly lower for these analyses. Differences were less clear for the office data as the differences between data sets was smaller.

This outcome suggests that screening symptoms that do not disappear or reduce after the work day is concluded may not be valid for this group of participants. The analyses showed that more evidence was provided to suggest that the lighting treatments differed (higher F Test values) when all symptoms were included, regardless of whether they continued or abated after the work shift. The model frequently adopted for studies examining the effect of the indoor environment on health assumes that when symptoms do not disappear or reduce after the work shift is completed they are due to non work related factors, typically illness such as influenza or seasonal allergies (hayfever) that are independent of the workplace (World Health Organisation, 1 984). If this was the case, then it would be expected that the additional data would decrease the difference between lighting treatments and reduce F Test values, as symptoms reported would be unrelated to workplace conditions. As the opposite appears to be the case, it suggests that valuable information may be discarded by screening the data for symptoms that remain once the work environment is left.

It is possible that this outcome was because the study population were shift workers. Many of the office personnel were also involved in tertiary study, were homemakers (many with children) or had other part time or even full time work. Therefore these participants are likely to be fatigued, leading to increased symptom incidence and/or severity, which may have influenced the prevalence at the end of the work day. In addition, office personnel who work an evening shift are likely to go home and shortly thereafter to bed. There may not be a sufficient period of time for symptoms to reduce or disappear. Therefore, this method of analysis may not be appropriate for shift workers.

Alternatively, symptoms caused or influenced by the lighting conditions studied in this research may not disappear or reduce after the work shift is concluded. If the symptoms are due to visual fatigue, and have a physiological basis, then the symptoms may not abate until the eye is rested by sleep. This is supported by research on fluorescent light flicker that reveals underlying physiological mechanisms theorised to cause differences in visual performance, comfort and fatigue under light

of differing frequencies (Eysel & Burandt, 1 984; Kennedy & Murray, 1 99 1 ). Further,

the study by Veitch & Newsham (1 998a) showed that visual fatigue could be

measured after exposure to differing flicker frequencies. This study did not show differences due to fluorescent light flicker, however the mechanisms may be comparable. Therefore in this experiment, the data sets that include the complete data set are likely to provide a better basis for assessing lighting treatment differences.

2. Treating the baseline, age and gender data as covariates;

The analyses considered the influence that office, age and gender differences had on symptoms reported by participants. When treatment differences were considered, all analyses had comparable trends to the other analyses undertaken, but with lower F test and corresponding p-values. The covariate data was not available for all participants therefore a smaller number of data points were available for analyses. Fewer data points reduces the statistical power of the analyses and inflates the probability of a Type 11 ((3) error, decreasing the probability of detecting any effect that may be present.

Therefore, as the baseline, age and gender balance across the three offices appeared to have minimal impact on the ANOV A outcomes and a smaller data set was used for the analyses it is probable that the power of the study was reduced and that differences between the offices due to these factors is unlikely to have been large or very influential.

3. Excluding responses in which the participants had no symptoms.

These analyses excluded responses where symptoms were not experienced. The

ANOV A provided strong evidence to show that eye and lethargy symptom severity differed between lighting treatments. In the eye symptom analyses, F test values were higher than in the data sets that included 'no symptom' responses, showing that the difference between lighting treatments could largely be attributed to participants experiencing more severe symptoms in the low frequency halophosphate lighting treatment. In the lethargy symptom analyses the lower F test value indicated that both symptom incidence and severity differed between lighting treatments.

These results suggest that participants who experience symptoms frequently in the workplace, were more sensitive to changes in lighting conditions than those who rarely or never report symptoms. Participants experiencing eyestrain and lethargy symptoms, reported that these symptoms were more severe in the low frequency halophosphate lighting treatment, in comparison to the low frequency triphosphor or high frequency triphosphor lighting conditions60•

Extrem e Values

Thirteen individuals reported outlying responses across the three trials with seven participants reporting severe symptoms in only one trial, five participants experiencing severe symptoms in two trials and one participant experiencing severe symptoms in all three trials. No relationship appeared to be present between the trials or the lighting treatments in which participants experienced symptoms.

60 More than 35% of participants reported severe eyestrain, headache or lethargy symptoms across the three trials (symptom severity of 5 or over). Therefore there was no evidence to suggest that a small group of participants were influential.

Seven of these individuals took part in the medical study61 and were found to have the following: insomnia (2), poor vision ( 1 ), back strain (5), hayfever (2), asthma (2). In four cases the results from the medical study strongly suggested that symptoms could be attributed in part to these factors. In particular, the participant who experienced severe symptoms in all three trials had very poor vision and her symptoms are likely to be solely attributed to this. Interestingly, she did not experience severe eyestrain, but did report severe headaches and concentration difficulties. The fmal three participants who took part in the medical study experienced symptoms that could not be attributed to factors outside of the work environment.

The participants who experienced extreme symptoms did not appear to be exceptional. Many symptoms experienced by participants could be explained (at least in part) by external factors or medical conditions, however in up to one third of cases, symptoms could not be adequately explained by these factors alone62• However, given the influential nature of their responses, which constituted only a small proportion of the total data points, examining the results without these data point included was appropriate.