Prevalence of eye and visual symptoms among office
workers and their relationship to self-assessed
productivity loss
Josefin Wärme
20
2020
2020
Student thesis, advanced level (master’s degree), 30 hp Occupational Health
Master programme in Health at Work Master thesis in Occupational Health Science
Supervisor: Hans Richter Examiner: Fredrik Hellström
Abstract
Josefin Wärme, Visual symptoms among office workers and their relationship to productivity loss, Student Master Thesis in Occupational Health Science, University of Gävle.
Aim: The dual aims of this descriptive cross-sectional study were: 1) to assess the
prevalence of eye-and visual symptoms among a population of office workers and; 2) to analyse if these symptoms were associated with self-assessed productivity.
Method: A questionnaire consisting of the Computer Vision Syndrome Questionnaire,
Convergence Insufficiency Symptom Scale and the Work Limitations Questionnaire was provided to each employee. Descriptive statistics on the number of individuals
classified as eye-and visual symptom cases was computed. Multiple logistic regressions analyses were performed on the individual eye-and visual symptom scores as
independent variables with the self-assessed productivity limitation scores as the dependent variable.
Main result: Out of 127 office workers, 76 answered the questionnaire (60% response
rate). The estimated prevalence’s of eye- and visual symptoms were 73% (95% CI: 61−83%) for Computer Vision Syndrome and 32% (95% CI: 21−43%) for Convergence Insufficiency-related symptoms. The multiple regression analyses revealed strong positive associations between eye/visual symptoms and productivity limitations for both Computer Vision Syndrome scores (p<0.001, r2=0,22) and Convergence Insufficiency-related symptoms scores (p<0.001, r2=0,39).
Conclusions: Symptoms of Computer Vision Syndrome and Convergence Insufficiency
were both prevalent, the former more so than the later. These symptoms were both strongly associated with limitations in self-assessed productivity. More research efforts are warranted to replicate and explore these work and health associations.
Sammanfattning
Josefin Wärme, Visuella symptom och dess samband med förlust av arbetsproduktivitet bland kontorsarbetare, Examensarbete i arbetshälsovetenskap Student, Högskolan i Gävle.
Syfte: Det första syftet i studien var att uppskatta prevalensen av Computer Vision
Syndrome och symtom relaterade till konvergensinsufficiens i en population av kontorsarbetare. Det andra syftet var att analysera om det fanns ett samband mellan dessa symtom och självskattad produktivitetsförlust eller inte.
Metod: En kvantitativ enkätstudie med tvärsnittsdesign genomfördes med följande
enkätinstrument: Computer Vision Syndrome Questionnaire, Convergence Insufficiency Symptom Scale and Work Limitations Questionnaire. Deskriptiv statistik beräknades för prevalenserna av antalet fall som var drabbade av respektive arbetshälsoproblem. Två separata multipla regressionsanalyser genomfördes dessutom.
Huvudresultat: 76 kontorsarbetare svarade på enkäten (60 % svarsfrekvens). Den
uppskattade prevalensen av Computer Vision Syndrome var 73% (95% CI: 61−83%) och av symtom relaterade till konvergensinsufficiens var 32% (95% CI: 21−43%). De multipla regressionsanalyserna visade statistiskt signifikanta (p < 0,001) samband mellan både Computer Vision Syndrome och symtom relaterade till
konvergensinsufficiens, med självskattad produktivitetsförlust. Ju mer synsymtom en person skattade, desto högre var den självskattade produktivitetsförlusten.
Slutsats: Nästan tre fjärdedelar av kontorsarbetarna hade Computer Vision Syndrome
och en tredjedel uppgav symtom relaterade till konvergensinsufficiens. Detta är ett arbetshälsoproblem för de drabbade kontorsarbetarna som behöver åtgärdas, både för de möjliga arbetshälsovinsterna för arbetstagarna och för de möjliga
produktivitetsvinsterna för företagen. Större studier behövs för att bekräfta och vidare analysera sambanden mellan synhälsa och arbetsproduktivitet.
Nyckelord: visuella symptom, kontorsarbetare, arbetsproduktivitet, computer vision syndrome, konvergensinsufficiens.
Preface
What a fantastic journey this paper has been for me. I have learned so much and for this I am forever grateful. My gratitude goes out to the Swedish education system, which is generous enough to provide me with all this knowledge I now carry, free of charge. That is truly a great privilege.
I also want to thank some important people who have helped and facilitated the creation of this paper. First, I want to direct my heartful thanks to my wise and talented
supervisor, Hans Richter, for patiently guiding me through my own winding thoughts. Thank you for keeping my focus on the main issue every step of the way. I could not ask for a better supervisor.
Second, I would like to thank my fellow students on the master program in occupational health, Emelie, Pernilla, Elin, Ivana and Natalie. Making this journey alongside such intelligent and inspiring women made it a whole lot more fun than it otherwise would have been. Thank you for all your valuable input on my paper and work process. Third, I want to thank Malin, my contact at the company from which I recruited my study population. Thank you for all your help and for remaining through the data collection despite of the consequences of the Covid-19 pandemic. I also want to thank all study participants who took the time and energy to answer my survey. Without you I would never have been able to complete this paper.
Fourth, I would like to thank my dear friends, Lisa and Hanna, who took the time to read my paper and give their wise criticism.
Last, the people I can never thank enough, my fantastic partner, Nicklas, and our
Index
Background ... 1
Near computer work ... 1
CVS ... 1
Convergence insufficiency ... 2
Relevance to health ... 3
Work productivity ... 4
Problem formulation ... 6
Aim and research questions/hypothesis ... 7
Method ... 7
Study design ... 7
Sample ... 8
Literature search ... 8
Data collection and surveys ... 10
Statistical analysis ... 13
Ethical considerations ... 19
Result ... 19
Background questions ... 20
Prevalence of CVS and CI-related symptoms ... 20
Productivity loss ... 21
Control analysis for effect of lay-off period ... 23
Multiple regression analyses ... 24
Discussion ... 27
Result discussion ... 27
Method discussion ... 32
Conclusion ... 36 References
Appendix 1 - Information letter to organisations or companies Appendix 2 - Search terms for the literature search
Appendix 4 - Quality assessment form for observational cohort and cross-sectional studies
Appendix 5 - Quality assessment form for case-control studies
Appendix 6 - Quality assessment form for controlled intervention studies
Appendix 7 - Quality assessment form for systematic reviews and meta-analyses Appendix 8 - Article matrix for background
1
Background
Near computer work
The technological development in computer science and communication has led to a transformation in the way people work. Seventy one percent of the employed or self-employed population in Sweden use computer devices and internet in their daily work (1). The visual function of the eyes is essential for these types of professions. It is however common for computer workers to experience symptoms linked to vision and eyes (2).
CVS
Computer vision syndrome (CVS), sometimes referred to as Digital Eye Strain, refers to visual and eye symptoms that affects some people after reading written material
displayed electronically on some kind of IT application (e.g. tablet, smartphone, stationary computer) (2). The symptoms most commonly associated with CVS are “eyestrain”, “headaches”, “blurred vision”, “dry eyes” and “neck and shoulder pain”. Why some people develop these symptoms is not completely understood. According to some researchers the eyes must work harder when reading text displayed on IT
applications compared to traditional forms such as printed paper. One study showed that participants reported more visual symptoms after reading a text from a computer screen compared to from a hard copy (3). The visual symptoms can become more pronounced if the person has uncorrected visual problems (2).
2
object in single vision. The projection of the image must remain in the centre of the retina for both eyes to maintain a single vision of the object. The eyes rotates towards each other (converge) when an object at near is fixated and rotates apart from each other (diverge) when an object at far is viewed (6, 7).
The prevalence of CVS in the general population is difficult to determine due to lack of standardized definitions and usage of the term. One review reports the prevalence of CVS to be at least 50%, but possibly more, in the general population (8). Another review reports the prevalence of CVS among computer users to range between 64 and 90% (4).
Convergence insufficiency
Employees suffering from binocular vision related dysfunctions develop visual symptoms when performing work tasks requiring near vision (8). Convergence
insufficiency (CI) or convergence disorder is a sensory and neuromuscular anomaly of the binocular vision system, characterized by a reduced ability of the eyes to turn towards each other, or sustain convergence. Convergence insufficiency is a condition which only manifests itself when doing near work, such as reading (9). When doing near vision work, a person with CI will have problems keeping the two eyes working together, that is, maintain the binocular function. A person with CI will experience symptoms when doing near work, such as “headaches”, “intermittent blur”,
“intermittent diplopia” (double vision), , “burning eye sensation”, “tearing”, “inability to keep up and concentrate at near”, “a sensation of words moving on the page”, “sleepiness when reading”, “decreased reading comprehension” and/or an “experience of reading slowly” (10). These symptoms are more prevalent with mental fatigue. The prevalence of CI, among other accommodative and non-strabismic (not related to squint) binocular dysfunctions, was researched in a review by Cacho-Martínez, García-Muñoz and Ruiz-Cantero (10). They emphasized the difficulties in presenting a
3
criteria that was used in the studies, the lower the prevalence of CI was. A range of prevalence was however presented at somewhere between 2,25-33%. Only three out of ten studies reviewed were done on adults and the prevalence range for those studies were 3,5-14%. Another review reports a prevalence of CI between 2-17% in the general population (11). The review also conclude that the prevalence is higher with increasing age.
Relevance to health
To clarify the relevance of vision related problems to health, the theoretical framework of the stress-vulnerability model was used. The model was originally constructed to explain the relationship of stress and vulnerability in relation to schizophrenia (12), but has been developed to be applicable to different types of psychological health (13). According to the model all people have a certain degree of vulnerability (12). The vulnerability has both an internal part, which is genetic, and an acquired part, which people get through life experiences. The stress factor consists of different types of stressors that challenge the individual. The relationship between vulnerability, stress and health/unhealth are shown in figure 1 below.
Figure 1. The stress-vulnerability model visualised.
4
level of the individual, is below the threshold, the individual has a natural response to the stressors in a healthy manner (12).
A person who experience a higher cognitive load reasonably do not perform as well as a person without excessive cognitive load, as cognitive resources not linked to the work task are allocated to deal with the CI-related symptoms (14). Research shows that feeling comfortable at work has an influence on work productivity (15). Well-being and task performance are two outcomes which are closely intertwined and one can influence the other (16). If there is a misfit between a person’s capabilities and the demands of the environment, reduced well-being and performance may occur.
Work productivity
Work productivity refers to the size of the production value, during a given time period, in relation to the size of the work effort during the same period of time (17). In the present study, the work productivity is applied to the individual worker and what is measured is the self-rated loss in productivity, due to vision problems, compared to the workers own assessed work productivity capacity.
There are some speculations concerning the effect of CVS on work productivity in the literature. One review makes the assumption that CVS affects occupational productivity based on the high symptom prevalence (4). When searching the literature, only one article was found which measured symptoms of CVS and self-rated productivity (18). It showed that there were no association between self-rated work performance and visual symptoms.
Astigmatism is another optical vision problem which can generate symptoms of CVS during computer work. Astigmatism is an optical problem where the light which passes through the cornea creates multiple focus points instead of one (19). This can be caused by an irregular cornea or lens. One study has researched the effect of induced
5
impact dry eye has on work performance (21, 22, 23). All three studies showed that the work productivity was lower among participants with dry eye.
No study has researched if presence of CI-related symptoms influences work productivity. There are however many studies that have investigated the association between presence of CI and reading performance or academic performance (24, 25, 26, 27, 28). Three studies measured ADHD behaviours among school children with
symptomatic CI (24, 25, 26). Children with presence of CI had significantly more ADHD-related behaviours during school activities, such as reading, compared to children with normal binocular vision (24, 26). One study showed that if CI was treated successfully, the ADHD-related behaviours significantly decreased among the children (25). Another study, which investigated two different treatments of CI among children with CI, found that both reading time and number of errors was significantly improved after completing treatment, compared to the control group (27). One study found no association between presence of CI and reading accuracy and reading comprehension (28).
Although all studies found, on CI and productivity, were done on children, there are reasons to assume that presence of CI-related symptoms could impact the work performance of adults as well. There is a thoroughly researched questionnaire which measures symptoms of CI called the Convergence Insufficiency Symptom Survey (CISS) (29). It has excellent validity and reliability, when tested on adults, and can be considered to work as golden standard for research concerning subjective symptoms of CI. The questionnaire entails, among others, the following questions: “Do you have trouble remembering what you have read?”, “Do you feel like you read slowly?”, “Do you lose your place when reading or doing close work?” and “Do you have to re-read the same line of words when reading?”. These questions, on face validity alone, suggests that the work performance can be affected if the respondent experience CI-related symptoms when doing near work.
New research indicates that there may exist a connection between CI and taxed
6
Binocular and accommodative anomalies may cause some employees to avoid or to reduce demanding near work and thus report no symptoms (31). To avoid or reduce the amount of near work may be the most extreme consequence of binocular and
accommodative anomalies, from a work productivity point of view. If a person is inhibited to perform at work due to vision problems, this circumstance could contribute to a sense of alienation, low self-esteem and reduced well-being.
In sum, it is likely that work performance and productivity could be influenced by CI. The consequences of this relatively common visual problem are currently unclear for work life performance, productivity, comfort and health.
Problem formulation
This study contributes to occupational health science in the following related ways: There is a lack of recent prevalence studies conducted, which have used validated questionnaires to measure CVS and CI-related symptoms, in Swedish office work. This study will investigate how common these visual problems are among a population of Swedish office workers. This will help to get a sense of the magnitude of the problem in this population, which will be valuable information for future research concerning vision problems among computer workers.
Information about the magnitude of the problem can be valuable for those who are responsible for the health of employees. Since one of the most common reasons for CVS is uncorrected or wrongly corrected vision for near work, one easy and
inexpensive intervention, which might be initiated after the results of this study is made known to the organization, is prescription and usage of new eye-glasses or so called “computer eye-glasses”, for those who are in need.
If CVS and CI-related symptoms among office workers are associated with reduced work productivity, it is a problem both for the office worker and for the employer. To be unable to perform fully at work, combined with experience of visual symptoms is likely to reduce the well-being and quality of life of the office worker.
The reduced work productivity is relevant to the employer as it will affect the
7
productivity and no study has done so using validated measurements. It is important not to limit research to only focus on the health effects of insufficient visual ergonomics, but also focus on the productivity aspects (16). Dul et al. emphasizes that the
stakeholders, that is the employers, with power to improve the work environment, will only do so if there are money to be saved in the process.
CVS can be prevented by ensuring optimal visual ergonomic work conditions and by using individualized vision aid (computer glasses or lenses) (2). CI can be diagnosed and managed (25, 27). If the productivity can improve among the workers by preventing CI-related visual problems from occurring (by use of customised eyeglasses for near work) or by screening for CI and treat the effected workers, this will benefit both the worker and the employer. Employees should learn the importance of paying attention to eye health, how to prevent eye symptoms from occurring and what to do if eye concerns arise.
Aim and research questions/hypothesis
The aim is to assess the prevalence of CVS and CI-related symptoms, the association between such symptoms and self-assessed productivity loss, in office workers.
1. What is the prevalence of CVS and CI-related symptoms in the study population?
2. Is there an association between CVS and/or CI-related symptoms and self-assessed productivity loss?
Method
Study design
8
questionnaire study is an easy way of gathering data from many study participants at the same time and it is also easy to enter the gathered data into a statistical programme. The empirical study was preceded by a systematic literature search to support the objective and research questions of the study. The guidelines, which have been used to formulate the final paper, are the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) Statement: guidelines for reporting observational studies (33). The directions for a cross-sectional study were followed.
Sample
The study population were office workers who spent most of their workday in front of a computer screen. The sample was recruited from the administrative staff working at a large vehicle production company in Sweden. This was done through a convenience sample, which is a form of non-probability sampling where participants are recruited on the basis of eligibility (32, s. 243-244). A letter of invitation to take part in the study was sent to the production company (appendix 1). After agreeing to participate, the contact at the company was sent the invitation to participants along with the weblink to the survey (appendix 1). Next, the contact forwarded this invitation to the eligible participants. Last in appendix 1 is the consent form, which the participants had to accept before they could answer the web survey.
A total number of 127 eligible participants in 14 different departments were invited to participate. The majority of the participants worked in an office landscape and a few worked in offices shared by 2-4 people. The study population was a mixture of different administrative staff, such as human resources staff, economy staff, logistic planners, project managers, editor staff and product designers. Most of the participants worked in front of stationary computer screens and many of them used double screens. One or two worked in front of a laptop screen. All participants worked at hight adjustable desks. All employees worked in sufficient daylight, but the tables were not all placed so that the light fell in from the side. All participants had the possibility to shade off if they were bothered by sun reflexes on the computer screen or elsewhere.
Literature search
9
The literature searches, which generated some of the articles in the background, was conducted in the databases Web of Science, Scopus and Medline. Web of Science and Scopus was chosen for their wide range of topics and because they have previously generated a relatively large number of search results on this subject. Medline was chosen on account of the medical symptomatic aspect of CVS and CI. Two different searches were made in these three databases. The first search had a broad meaning of the internal exposure, se table 1 in appendix 2, for the search terms used in the first search. This search generated few relevant studies, demonstrating the limited research done on this subject, CVS or CI and their effect on work productivity. Another search was made with intention to broaden the research outcome, productivity or performance, to incorporate other types of near work exposures, such as reading and academic achievement. The search terms for the second search are presented in table 2 in appendix 2.
During the literature search in the databases the limitations where set to English
language, articles as document type and journals as source types. No limitation for peer reviewed articles was set with the aim to avoid removal of relevant articles due to technical error. The journals who published the relevant articles, which passed the practical screen later on, were reviewed manually by the author to confirm that the articles were peer reviewed. If no such confirmation could be made, the article was not to be included for quality screening. However, no relevant article was published in a journal that did not apply peer review before publishing.
To facilitate the process of the practical screen, the inclusion- and exclusion criteria used in each search (34) was compiled in questionnaires. The practical screen
questionnaire for the first search is found in the first table in appendix 3 and the one for the second search is found in the second table in appendix 3.
10
that were not deemed applicable for any of the assessed studies. Question nr. 8 in the assessment tool for observational cohort and cross-sectional studies was added by the author to save time when assessing cross-sectional studies (see appendix 4). Two questions were added at the end of each form to assess generalizability of the studies in the context of the intended study population for this study. All four forms used for quality assessment can be viewed in appendix 4-7.
The quality assessment forms scored one point for every “yes”. “No” and “unclear” as answers gave zero points. The levels of scoring were low, medium and high quality. The limits of scoring were set at a minimum of 80% for high quality and a minimum of 60% for medium quality. The percentage was calculated of all questions deemed applicable for each study. A question answered with “not applicable” reduced the total maximum score with 1 point. The percentage was rounded to the nearest integer point. For example, if the study was a case-control, and all questions were deemed applicable, the maximum point was 12. 80% of 12 is 9,6 (12 x 0,8) and when rounded off to the nearest integer, the limit for high quality was set at 10 points. The limit for medium quality was 7 points (12 x 0,6 = 7,2).
The articles found through the literature searches are presented in a matrix in appendix 8, along with other articles which helped build the background.
Data collection and surveys
Invitations to participate in the study were sent to the participants on March 13th and April 28th, 2020. There was an unplanned pause in the data collection due to the Covid-19 pandemic which caused the vehicle company to lay off its staff for five weeks. About half of the data was collected in March and the other half in late April. A questionnaire was distributed to the intended participants, which they were instructed to complete individually. The questionnaire was constructed in the survey tool Sunet Survey (36) and comprised of 50 questions in total. A weblink to the online survey was sent by e-mail to the eligible participants and they had to give their consent to participate before they could answer the survey. A reminder to complete the questionnaire was sent out five respectively one week/weeks after the first send-out, depending on if the
11 Background- and control questions
Questions related to demographic data, such as age, gender and earlier eye/vision diagnosis, along with possible confounding factors were inserted in to the first part of the survey, see appendix 9. The control questions concern general health, sleep quality, how many hours per day the workers estimated that they spent in front of a digital screen at work respectively in their free time, if they used glasses or lenses and how long ago they had their most recent eye examination. One question asked participants to assess their general health, this question was retrieved from the Swedish version of the health survey SF-36 (37). The reason for including general health as a possible
confounding factor was that the CI-related symptoms could possibly be misinterpreted symptoms of burnout. As burnout includes symptoms of cognitive impairment such as difficulties to concentrate (38), and when working in front of a computer screen it could be hard to distinguish between concentration difficulties due to burnout and those due to eye symptoms. The last of the control questions in appendix 9 concern sleep quality and was borrowed from the Swedish version of the Karolinska Sleep Questionnaire (KSQ) (39, 40). Sleep quality was included as a confounding factor as previous research has shown that it can impact a person’s productivity (41, 42, 43).
Computer Vision Syndrome Questionnaire (CVS-Q)
To measure the office workers self-rated symptoms of CVS a validated questionnaire called Computer Vision Syndrome Questionnaire (CVS-Q) was used (44). It contains 16 different symptoms which the respondent was asked to rate in both frequency and intensity. The frequencies were never (0 p), occasionally (1 p) and often/always (2 p) and intensity was rated as moderate (1 p) or intense (2 p). The two ratings were then multiplicated with each other to present a single score for each symptom. Then all symptom scores were summarized, and it was this sum of all scores that was used in the statistical analyses. A summarized symptom score of > 6 suggests the presence of CVS (44). The maximum score that can be achieved with the CVS-Q is 64 points.
12
questionnaire have consulted experts and scientists in the field with regard of the questionnaire´s ability to measure the syndrome and, according to them, there was a wide consensus about the questionnaire’s ability to do so (44). With good reliability and validity within the population of computer workers, the questionnaire is deemed to be a trustworthy instrument to measure CVS. In this study the questionnaire was translated to Swedish by the author, see appendix 9.
Convergence Insufficiency Symptom Survey (CISS)
To diagnose CI in patients, an examination of the eyes must be done by a professional practitioner (11). CI is usually diagnosed by a near point of convergence (NPC) which is receded, a greater exodeviation (latent outwardly strabismus) at near then at far and reduced convergence amplitudes at near. No measurements to diagnose CI were possible in this study, as it is a survey study. There is a self-assessed screening tool called Convergence Insufficiency Symptom Survey (CISS) which can measure CI-related symptoms and can be a good indicator of who is likely to suffer from CI (29). The CISS has good discrimination as the area under the ROC curve is 0,98. It has good to excellent internal consistency, with a Cronbach’s alpha of 0,96, and it has good to excellent reliability, with ICC of 0,88. The CISS has previously shown a wide spread in mean scores between people with normal binocular vision (NBV) and those with CI. As the CISS is a valid and reliable measurement, a Swedish version (translated by Tony Pansell, Karolinska Institutet, Lector/Optician) was used to identify participants with CI-related symptoms in this study. The CISS comprises of 15 questions which are answered on a 5-point Likert scale. The possible answers to each question are never (0 p), infrequently/not very often (1 p), sometimes (2 p), fairly often (3 p) or always (4 p). The scores for all questions are summarized and the maximum total score achievable is 60 points. For adults, a total score of > 21 is suggestive of presence of CI. The Swedish version of the CISS is presented in appendix 9.
Work Limitations Questionnaire (WLQ)
13
demands and output demands. These capture universal demands, some of which, according to the hypothesis, may be impacted by CVS and/or CI related problems. The questions are answered on a 5-point Likert scale. For each limitation scale an average item score is calculated (maximum 5 points for each scale) and converted to a to a 0-100 range (46). An algorithm is used to calculate a total WLQ Index, and finally a total percentage of loss in productivity the last two weeks. In the current study, the individual estimates of percentage loss in productivity were used in the statistical analyses. The highest loss of productivity a person can score with the WLQ is 25% (45). The highest summarized score possible for the WLQ is 125 points. Within both patient and
employee populations, the WLQ has shown good construct and criterion validity, and the scales are significantly related to the SF-36 measures of physical and mental health. Due to a User Licence Agreement (No. 238832) with Mapi Research Trust, the survey questions for this licenced instrument cannot be disclosed and are therefore left out of the survey presentation in appendix 9. A version of the questionnaire in Swedish was supplied by the questionnaire provider Mapi Research Trust (47).
Trial send-out
A trial send-out of the survey, from Sunet Survey, was conducted prior to the real data collection. The web link worked as intended and the survey took about 10-15 minutes to complete. Minor flaws were corrected. Most questions were answered with ratio buttons pre-coded to specific values. The answers were coded so the results could be transferred directly to SPSS, where the statistical analyses were conducted.
Statistical analysis
The statistical analysis was conducted with the statistical tool IBM SPSS Statistics Version 24.0 for Windows (48). In all analyses the level of significance were set at p < 0,05. All the scores from the three questionnaires, CISS, CVS-Q and WLQ, was treated as ratio/interval variables in the analyses in SPSS, as they were all numeric (32).
Research question 1: What is the prevalence of CVS and CI-related symptoms in the study population?
14
(total CISS score >21) or a CVS-case (total CVS-Q score >6). Presence of CVS and CI-related symptoms was, for this research question, treated as ordinal variables and was coded 0 = not a case and 1 = a case. The total number of cases and percentages thereof are presented numerically and graphically and presented separately for women and males.
What is the prevalence of estimated productivity loss in the study population?
Descriptive data of estimated productivity loss in the study population was computed using the WLQ algorithm which converted WLQ scale scores into an estimate of productivity (% of 100). The central tendency of the sampled group responses, in the categories: symptom free, CVS cases and CI cases, are presented numerically in the text and visually in comparing box-plots.
To control for any differences in productivity loss for the participants who answered the questionnaire before, respectively after, the lay off period, an unpaired t-test was
performed.
Research question 2: Is there an association between CVS and/or CI-related symptoms and self-assessed productivity loss?
Null hypothesis: there is no association between CVS or CI-related symptoms and self-assessed productivity loss.
Alternative hypothesis: there is an association between CVS or CI-related symptoms and self-assessed productivity loss.
The independent variables in this statistical analysis were CVS or CI-related symptoms and the dependent variable was self-assessed productivity loss. The total score of the CVS-Q or the CISS was used for the independent variable, which are numeric variables. The score for the self-assessed productivity loss was also treated as a numeric variable in the analysis.
15
variable, self-rated productivity loss, was a continuous numerical variable and it was appropriate to control for possible confounding factors. For these reasons, the multiple regression analysis (MRA) was chosen as method of analysis.
Two separate MRA models were built in the first analytic step. One with CVS as independent variable and self-assessed productivity loss as dependent variable, and one with CI-related symptoms as independent variable and self-assessed productivity loss as dependent variable.
To decide which confounding factors that could have any influence on the self-assessed productivity loss, a Pearson correlation analysis was made with all the possible
confounding factors from appendix 8 included. The analysis showed that, besides CVS and CI-related symptoms, the variables general health and sleep quality correlated significantly (0,481, p = 0,000, respectively 0,494 p = 0,000) with self-assessed productivity loss. Both general health and sleep quality were included in the MRA models. One rule of thumb, when it comes to the number of explanatory variables in the models, states that the number of explanatory variables should approximate the number of participants divided by 10 (49, p. 85). The MRA models included a minimum of 50 participants (for CVS, the regressions with CI-related symptoms included 59
participants), which means the model admits 5 explanatory variables in the analysis (50/10 = 5). As the regression analyses included 3 explanatory variables at a time, there was no risk of overfitting the models (49, p. 99).
Four symptoms rated were identical in the CVS-Q and CISS questionnaires (eye pain, blurred vision, double vision and headache). The similarities between CVS and CI-related symptoms were controlled for statistically by checking for collinearity. The Pearson correlation analysis showed that, with a value of 0,6 (correlation coefficient should be <0,8) (49, p. 100), CI-related symptoms and CVS did not correlate too strongly with each other. This is a testament that the instruments measures two separate vision related symptoms.
In the second analytic step, after the first two MRA models were performed (which included CVS and CI-related symptoms in separate models), two additional regression models were built which included only the significant variables passed on from
16
associations with productivity loss in the regression model for CVS, they were not included in the final MRA. Sleep quality had significant associations with productivity loss in the regression model with CI-related symptoms and was therefore kept as an explanatory variable in the final MRA model.
Checking the assumptions for multiple regression analysis
In order to perform a MRA there are several assumptions which need to be met (49). First, the relationship between the independent and the dependent variable need to be linear. This was checked by plotting the dependent variable against the independent variable and making sure the data approximated a straight line and by plotting the residuals against the independent variable, where a random scatter of plots was wanted. (Second, there can not be more than one pair of observations on each participant.) Third, the residuals should have a mean of 0 and be normally distributed. This was checked by producing a histogram and performing a test of normality. Fourth, the residuals should have a constant variance for all predicted values of the dependent variable. Fifth, the independent variable was measured without error. And sixth, the independent variables in the MRA can not be highly correlated with each other or multicollinear (50). To check this later assumption, a Pearson correlation matrix was produced for all the independent variables. A correlation coefficient between the variables should be <0,8. It was also checked by making sure the Variance Inflation Factor (VIF) values in the MRA were <10.
When checking the above mentioned assumptions, they were not originally met. The relationship between the dependent and the independent variables was not linear, thus breaking the first and most important assumption (49, p. 30). When performing a test of normality on the dependent variable and the independent variables, the dependent variable was not normally distributed. When viewing a histogram of the dependent variable it became clear the variable was positively skewed due to the fact that a large portion of the participants had scored 0 or close to 0% in productivity loss. Therefore, the dependent variable needed to be transformed. It was not sufficient to simply
17
of the regression analysis if CVS/CI-related symptoms are linearly associated with loss in productivity, as formulated in the research hypothesis. The linear relationship should still be present, even if the 0 values were deleted.
After removing all 0 values in the dependent variable, it was logarithmically
transformed to base 10 (log10 transformed) (49, p. 30). This transformation method was chosen because of the positively skewed distribution of the dependent variable and beacause there was an exponential relationship between the dependent variable and CVS or CI-related symptoms, see example in figure 2. A new test of normality (Kolmogorov-Smirnov) confirmed that the transformed variable was normally distributed. After the transformation, the relationship between the variables were approximately linearized, see figure 3, thus checking the first assumption. Assumption 2-4 were also controlled and checked for in all regressions. The fifth assumption, the independent variable should have been measured without error, was violated for CVS. This was dealt with as described under “Internal data loss”. CI-related symptoms and productivity loss was measured without error, in the sense that there were no missing values.
18
Figure 3. Productivity loss plotted against CI-related symptoms (CISS) after removing all zero values and logarithmically transforming the dependent variable to the base of 10. (WLQ = Work Limitation
Questionnaire).
Internal data loss
Forty (53%) of the surveys were returned with one or more missing response values in the CVS-Q. Thirty-one of the participants (41%) only provided values for either the intensity or the frequency of a given symptom category (i.e., failed to fill in the counterpart value for the specific symptom). For example, one participant might have answered that he/she experienced double vision “sometimes” (frequency) when working in front of a computer screen, but failed to fill in with which intensity he/she
experienced the symptom of double vision (moderate or intense). Eight of the participants (10%) omitted to reply to one full question (i.e., failed to fill in both the frequency and the intensity of the symptom). One participant (1%) skipped three questions in total in the CVS-Q.
19
of one (“occasionally” or “moderate”), which is equal to the lowest response alternative for presence of a symptom. Internal data loss occurred only for CVS. The surveys used for collecting CI-related symptoms and productivity loss did not contain any missing responses.
An additional correlation analysis was made with the variables, productivity loss in percent and total score of CVS among the 37 participants who answered the CVS-Q fully. This control analysis was made to see if a statistically significant correlation, between CVS and productivity loss, was present or not without the imputed values in the CVS-Q. The control analysis showed that there was a significant correlation.
Ethical considerations
The ethical considerations made in conducting this study were in agreement with the guidelines presented by the Swedish science council (51). The goal was to maintain the individual protection claim as far as possible. In accordance with the information claim, the study participants were, in writing, informed about the aim of the study, their role, what the participation would mean for them and how the data was collected and used. They were also informed that the participation was voluntary and that they, at any time, could withdraw from the study. The consent claim was met by informing participants that filling in the survey also meant giving consent to participate.
The confidentiality claim was met as all data was anonymized at collection (51). No answers could be connected to a specific individual, as no personal data was collected. The assurance that participants could not be identified in the data was important, as research concerning health is to be considered ethically delicate. All data was stored in a secure manner and the raw data, used for analysis, was stored on the hard drive of the University of Gävle. When the study is completed, all data will be deleted. The useful claim was met by not using or lending information about participants for commercial or other non-scientific use. The study was approved by the council of research ethics of Gävle University (2019-12-13, Forskningsetik Yttrande 2019_10).
Result
20
stated that they had an existing eye-/vision diagnosis and one person had skipped this question. To avoid this potentially confounding factor, these four people were excluded from analysis, making the final number of participants 76 office workers (60%).
Background questions
The mean age in the study population was 42 years (sd 9,5, min 24, max 66) and there were 31 women (41%) and 45 men (59%). The mean hours spent in front of a digital screen during work hours were 5,5 hours per day (sd 1,4, min 2, max 8) and, in spare time, 2 hours per day (sd 1,4, min 0, max 10). Fifty-eight participants (76%) used eyeglasses or lenses. For general health, 7 participants (9%) rated that they had excellent health, 34 (45%) had very good health, 31 (41%) had good health and 4
participants (5%) had moderate health. The mean for general health was 2,4 (sd 0,7, min 1, max 4), which is between very good and good. Regarding time passed since the last visit to an optician, 20 participants (26%) rated <6 months, 14 (18%) rated 6 months to 1 year, 20 (26%) rated 1-2 years, 12 (16%) rated 3-5 years and 10 participants (13%) rated that >5 years had passed. The mean for time passed was 2,7 (sd 1,4, min 1, max 5), which was between 6 months and 2 years. For general sleep quality, 18 participants (24%) rated that they slept very good, 43 (57%) slept fairly good, 9 (12%) slept neither good nor bad, 5 (7%) slept fairly bad and 1 participant (1%) slept very bad. The mean for general sleep quality was 2,0 (sd 0,9, min 1, max 5), which was fairly good.
Prevalence of CVS and CI-related symptoms
The prevalence of CVS was calculated from the 67 participants who had a total score in the CVS-Q after imputing values. The prevalence of CVS was 73% (95% confidence interval: 61−83%), which means that 49 participants, out of the 67, suffer from CVS. There was a minor difference in prevalence of CVS between women and men, where the prevalence for women was slightly higher, see table 3.
21
There were no missing values for the CISS questionnaire. Twenty-four out of the 76 study participants had CI-related symptoms, making the prevalence 32% (95%
confidence interval: 21−43%). The prevalence of CI-related symptoms for women was 29% and for men 33%, see table 4.
Table 4. Number and percentages of cases/no cases with/without CI-related symptoms for men respectively women.
Productivity loss
The median of productivity loss within the study population was 1,4% (95% confidence interval: 0,8−2,3%), with a slightly higher median value for men (1,6%) than for women (1,1%). The median for participants who did not have CVS was 0% (min 0, max 1,5) productivity loss and 2,3% (min 0,5, max 10,5) for those with CVS. The central
22
Figure 4. Box-plot for productivity loss for Computer Vision Syndrome-cases (CVS) respectively not cases. WLQ = Work Limitation Questionnaire.
The median for productivity loss for participants who were not classified as having CI-related symptoms was 0,5% (min 0, max 9,8) and 4,8% (min 0,5, max 10,5) for those with CI-related symptoms. The central tendency of self-estimated productivity loss for cases respectively no cases with/without CI-related symptoms are displayed in figure 5.
23
The medians and ranges for participants without both CVS and CI-related symptoms, for those with only CVS and those with both CVS and CI-related symptoms are displayed in table 5.
Table 5. Central tendency and spread of productivity loss in percent for participants without symptom presence, those with only CVS respectively those with both CVS and CI-related symptoms.
WLQ = Work Limitation Questionnaire, CVS = Computer Vision Syndrome, CI = Convergence Insufficiency (-related symptoms).
There were no participants who only had presence of CI-related symptoms, without also having CVS. However, there were four participants who had CI-related symptoms and missing values for CVS, but the participants without values for CVS are not included in table 5. Twenty participants were classified as both CVS and CI-related symptoms cases (see table 5), which is 30% of all participants and 41% of the participants who were classified with symptom presence (N = 49).
Control analysis for effect of lay-off period
There was, as described earlier, an unplanned pause in the data collection due to a five week long lay off period. Control analyses was conducted to find out if the lay-off period made any impact on the collected data set. A Wilcoxon rank sum test was made to compare if there were any differences in self-assessed productivity loss between the participants who answered the survey before respectively after the lay-off period. The p-value of the test was 0,96 (p > 0,05), with a Z statistic of -0,05. This means that there was no statistically significant difference in productivity loss between the ones who answered before and the ones who answered after the lay-off period.
24
Multiple regression analyses
CVS
In the first MRA with CVS, general health and sleep quality was controlled for. The model explained 35% (adjusted R square) of the variance in self-assessed productivity loss. The effect of CVS was significant (p-value <0,05), but the effect of general health and sleep quality was insignificant and therefore a new linear regression was performed with only CVS as independent variable. This second model explained 22% (adjusted R square) of the variance in self-assessed productivity loss. The result of this analysis is presented in table 6, where is it shown that CVS is significantly associated with self-assessed productivity loss with a p-value < 0,001.
Table 6. Coefficients table from MRA with CVS as independent variable and self-assessed productivity loss as dependent variable.
CVS = Computer Vision Syndrome. Significance level set at p < 0,05.
25
Figure 6. Computer Vision Syndrome (CVS) plotted against unstandardized predicted values of, log10 transformed, productivity loss. WLQ = Work Limitation Questionnaire.
The regression model predicts that if an office worker scores, for example, 15 in the CVSQ (a case), he or she will have a productivity loss of 0,496 (confidence interval -0,149 – 1,125) in a log10 transformed value (0,045 * 15 - 0,179 = 0,496). This means that the productivity loss is 1,6% (confidence interval 0,9 – 3,1%) for an office worker who scores 15 in the CVS-Q (e0,496 = 1,64). If an office worker scores 4 in the CVS-Q (not a case), he/she will have a productivity loss of 1% (confidence interval 0,7 – 1,5%). If he/she scores 24, the predicted productivity loss would be 2,5% (confidence interval 1,0 – 5,7%).
CISS
26
quality were significantly associated with self assessed productivity loss with p-values < 0,001.
Table 7. Coefficients table from MRA with CI-related symptoms as independent variable and self-assessed productivity loss as dependent variable.
CISS = Convergence Insufficiency Symptom Survey. Significance level set at p < 0,05.
The unstandardized B coefficient for CI-related symptoms is 0,022 (confidence interval 0,009-0,034), which is shown in table 7. For every one-unit increase in CISS, the productivity loss increases with 0,022, when sleep quality is fixed. The CISS total scores and sleep quality are ploted against the model predicted unstandardized log10 transformed values for productivity loss in figure 7. The 3D-plot allows us to visualize the 5 scale steps of the sleep quality variable, which generated different lines in the prediction model. For example, the line closest to the bottom of the plot are cases who rated their sleep quality as 1, very good.
27
According to this model, when calculating the predicted productivity loss for an office worker, his or her sleep quality should be taken into consideration. The mean for sleep quality was 2 (fairly good, sd 0,9) in the study population, therefore this will be used as the fixed value in the following examples. If a person scores 5 in total CISS score (not a case), he or she will, according to the model, have a productivity loss of 1% (0,022 * 5 + 0,225 * 2 - 0,58 = -0,02. e-0,02 = 1,0%), with a confidence interval of 0,5 – 1,8%. A person with 35 in CISS score (a case) will have 1,9% (CI 0,7 – 4,9%) in productivity loss (0,022 * 35 + 0,225 * 2 - 0,58 = 0,64. e0,64 = 1,9%).
The answer to the second research question, “is there an association between CVS and/or CI-related symptoms and self-assessed productivity loss?”, is that there is an association between both CVS and CI-related symptoms and self-assessed productivity loss. Thus, the null hypothesis can be rejected. The association is such that the higher an office worker scores in the CVS-Q or the CISS, the more reduced is his/her work
productivity.
Discussion
Result discussion
The aim of this study was to assess the prevalence of CVS and CI-related symptoms, the association between such symptoms and self-assessed productivity loss, in office workers.
Prevalence of CVS
28 Prevalence of CI-related symptoms
In the background, a review by Cacho-Martínez criticises the prevalence studies of CI for not being randomized (10). The study participants in this study were recruited through convenience sampling. This makes it difficult to draw any conclusions about the prevalence of CI in the population they came from. The review also noted that the more diagnostic criteria used in previous prevalence studies of CI, the lower the
prevalence percentage was (10). The prevalence of CI in earlier studies ranged between 2-33%. The prevalence in this study was 32%, which is a high prevalence in
comparison.
The high prevalence may be explained by the fact that the CISS questionnaire was the only measurement of CI-related symptoms. It is probable that the prevalence of CI would be lower if all the identified cases in this study were examined by an
ophthalmologist as well. But it would be rash to excuse the high prevalence of CI-related symptoms in this study as an error of measurement, as the fact is that 32% of the office workers experienced CI-related symptoms. That subjective experience cannot be ignored. Further, the study never aimed to assess the prevalence of CI, but CI-related symptoms. The symptoms of CI are there, and they were measured using the CISS questionnaire, which is considered a golden standard in predicting people which have CI. The high prevalence of CI-related symptoms indicates that this impairment of function might not be detected to a sufficient extent among office workers. This is a condition that can be detected through a relatively simple eye exam by either an optician or an ophthalmologist. There are methods to manage the symptoms and help affected people and, therefore, they should be identified.
29
It is possible that office workers with symptoms from the eyes were more likely to participate in this study than those without symptoms. Therefore, it is a possibility that the prevalence of CVS and CI-related symptoms are higher in the study population than in the true population of office workers. Another possibility is that the participants who experienced vision symptoms when doing near work avoid this sort of work as a coping strategy, thus lowering the prevalence of CVS and CI-related symptoms in the study. This coping mechanism is common among school children and adolescents with CI (26).
In this study no participants had presence of CI-related symptoms without having presence of CVS as well. This confirms the pre-existing notion that CI may be one of the probable causes of CVS (4).
Productivity loss
The mean of the self-assessed productivity loss, measured with the WLQ, was 2,6% in the study population. The mean for participants who were classified as CVS-cases was 3,3% and 4,7% for cases with CI-related symptoms. To put these percentages into perspective, they can be compared with percentages of productivity loss measured in other studies with the WLQ. One study on patients with rheumatoid arthritis (RA) showed a productivity loss of 7% (52). Another study among office workers in Japan reported a 5% loss in productivity among participants diagnosed with dry eye disease (DED) (23). The participants with CI-related symptoms had a self-assessed productivity loss comparable with people with diagnosed DED, but it is not as inhibiting as
diagnosed RA.
30
this example worth 23 500 SEK (500 000 SEK x 0,047) per year for that one employee (min 2 500 SEK, max 52 500 SEK). If the worker had CVS and instead had a
productivity loss of 3,3% (min 0,0%, max 10,5%) the lost work hours would be worth 16 500 SEK (min 0 SEK, max 52 500 SEK) per year.
Except for the obvious health gains for the employees, there are economical gains to be made to thoroughly investigate and take measures to improve the visual health of the office workers.
Regressions
There is an association between both CVS and CI-related symptoms, and self-assessed productivity loss and thus the null hypothesis for the second research question can be rejected. The result of the regression analyses in this study show that there were significant relationships between both CVS and self-assessed productivity loss and CI-related symptoms and self-assessed productivity loss. This result contradicts the previous study which investigated if there was an association between CVS and self-assessed work productivity and found no association (18). The previous study did not use standardized and validated questionnaires to measure neither CVS nor work productivity, which was done in this study.
The associations between the independent variables and the outcome were such, that the higher total score in CVS-Q or CISS a person has, the higher the self-rated productivity loss was. Sleep quality had a significant impact on the association between CI-related symptoms and productivity loss. The worse sleep quality a person rated, the higher the productivity loss, when the total CISS score was fixed.
The regression models explain 22% (CVS) respectively 39% (CI-related symptoms) of the variance in productivity loss. There is an error variance in both models, which cannot be explained with the prevalence of CVS or CI-related symptoms (the later combined with sleep quality). One possible explanation for the error variance could be that the experience of visual symptoms is a subjective one. Visual symptoms include different experiences of discomfort related to the vision. The subjectiveness of the symptoms may be comparable to the subjectiveness of pain. The International
31
experience is always individual and factors that influence the experience can be
biologically, psychologically and socially derived. This biopsychosocial outlook can be applied to vision symptoms as well. Two people with similar physical vision problems might experience the visual symptoms differently.
Theoretical perspectives
The Job Demands and Control (JDC) model by Karasek and Theorell (55, 56) has, originally, two dimensions: job demands and control. According to this model, psychological strain occurs when job demands are high, and the control or decision latitude is low. Job demands is the joint effects from all stressors in a job situation and control is the workers’ freedom to control his/her activities and use of skills. For the office worker, the screen time constitutes a demand in the work environment. The office worker needs to work in front of a computer screen in his or her work and this requires normal binocular vision. A person who experience vision symptoms, due to CVS or CI, while working in front of a computer screen becomes hindered in utilizing their skills to full extent.
The stress-vulnerability model, which was explained in the background, emphasizes the additive effect of the vulnerability traits (12). The vulnerability is the sum of the
individuals’ inborn and acquired vulnerability. The acquired vulnerability covers both psychological and social aspects which can influence the persons’ proneness to have an unhealthy response to stressors in the environment. Vision symptoms, due to CVS or CI, in themselves are not likely to cause an unhealthy stress reaction. There is, however, a risk that the additive effect of the vision symptoms, on other vulnerability factors, could contribute to stress-related ill health, such as burnout. The risk for this is higher if the individual has a stressful work environment, as the cognitive load is reasonably higher for a person with vision symptoms than for a person without, given that their stressors and vulnerability levels are otherwise the same.
Contribution of knowledge
32
office workers suffering from CVS and CI-related symptoms are likely to have
diminished work productivity and the more symptoms they have the more affected their productivity is. The result of this study might motivate the leaders of companies with office workers to investigate their employees’ visual health status and help the affected workers get the help that they need.
This study also contributes to science as it presents prevalence numbers of CVS in a population of Swedish office workers, which no other study has done before with a validated instrument of measurement. There are no prevalence studies of CI-related symptoms among adults done in a population of Swedish office workers, which is also a scientific contribution of this study. The study is also the first to show that there are associations between CVS and CI-related symptoms and self-assessed productivity loss. This is valuable to science as there now is a clear indication that the work productivity can be inhibited by these visual symptoms. Earlier this association has only been presumed or not been captured.
Method discussion
The cross sectional study design is a limitation of this study as the data was collected at one fixed point in time and therefore it is not possible to distinguish if one variable causes the other (32, p. 87-88). In this study, it is not possible to conclude that vision symptoms cause productivity loss as the two variables were measured at the same time. What can be concluded is the prevalence of CVS and CI-related symptoms and the associations with productivity loss that was registered at the point of the data collection. However, the questions in the WLQ asks how much the person is hindered to perform different tasks due to vision problems. This way of asking questions suggests that the vision problems should come before the loss in productivity, but longitudinal studies are needed to establish a causal relationship between the variables.
33
eligible participants. The generalizability of the study results is limited by the fact that convenience sampling was used (32). The results cannot be generalized to all office workers who work in front of a computer screen, as it is probable that the study
population is not representative for the true population. It is, for example, possible that the office workers who experience vision problems were more likely to participate as they could relate to the aim of the study.
One strength in this study is that the response rate was higher than anticipated at 63%, as response rates to e-mail surveys can be as low as 20-30% (58). This increases the chance that the study population is representative for the group of office workers who were invited to participate in the study. One reason for the high response rate might be that the invitation to participants was sent from the contact at the company. When the e-mail came from an address within the company itself, the study might have been
perceived as less foreign and as something more related to the company’s own interests. The number of participants in this study was 76 in total and the regression analyses were made on fewer participants (50 for CVS and 59 for CI-related symptoms) related to missing values in CVS-Q and removal of zero values in the WLQ. These are
relatively few participants, which is a limitation of the study, but despite of this the results of the analyses showed statistically significant associations. This can indicate strong associations between the variables, but it cannot be ruled out that the associations arose by chance, as there is a 5% risk that the results of the analysis came by chance. Studies with more participants, preferably sampled through randomization, are called for to confirm the associations.
The confidence intervals, for the predicted self-assessed productivity loss, in both MRAs were relatively large. In the example given in the result, with a person who had a total CISS score of 35 (and rated his/her sleep quality as fairly good), the true
productivity loss could be between 0,7 and 4,9%. More studies with larger study populations are needed to determine the slope of the association between CI-related symptoms and productivity loss.
34
of the study design. It can be hard to motivate the benefit of treating symptoms to employers, if the benefits in doing so are abstract and subjective (16). In this study the value of treating affected employees is emphasized by measuring productivity loss and by showing that there is an association between vision symptoms and productivity loss. There might be economical gains to be made and this is clarified in this study. However, larger studies are needed to establish the worth of the economic gains.
The fact that the data collection was interrupted, when all the office workers were laid off for 5 weeks due to the Covid-19 pandemic, might have had an impact on the study result. About half of the data was collected before the lay off period and the other half after the lay off period. Both the CVS-Q and the CISS asked about vision symptoms without asking about a specific time period, the WLQ however asked the participant to rate the impact of his/her vision problems on his/her work in the last two weeks. It also specified that the questions should be answered even if the participant had missed a few workdays. The participants who answered the survey after the lay off period might have rated their productivity for a time period when they did not work. This is a limitation of the study. This was however statistically controlled for by performing an unpaired t-test, which showed that there was no difference in productivity loss between the participants who answered the questionnaire before the lay off period and the ones who answered it after. This is promising, but it is still possible that the lay off period affected the study results in some other way. Therefore, further studies that research the associations between vision problems and productivity loss are called for.
The internal data loss in the CVS-Q is another limitation of the study. The data loss was handled by imputing values for participants who had rated either frequency or intensity of a symptom without filling in the counterpart of the question. This imputation
35
There is a possibility of capturing burnout with the CISS, as the questionnaire asks questions about, for example, a person’s ability to keep focus when doing near work. A person, with burnout tendencies, would likely score higher on these questions, then a person without burnout tendencies, if their binocular vision was similar. Burnout was not controlled for in this study, as this would have required an additional battery of questions, which there was not room for in the survey. The survey already compiled of 50 questions and there was a risk of losing respondents if the survey became too time- and energy consuming. The compromise made was to instead control for general health with one question borrowed from the health survey SF-36 (37). General health
correlated significantly with productivity loss, but the association was not significant in the MRA models with CVS and CI-related symptoms. In other words, the general health of the participants had no statistically significant association with self-rated productivity loss.
Reliability and validity
The use of validated instruments of measurements (CVS-Q, CISS and WLQ) is a strength in this study (44, 29, 45). With high validity and specificity there is a limited risk of being misclassified as a case/not a case. All three questionnaires are the some of the best ways of measuring CVS, CI and productivity loss, through self-rating,
available.
The WLQ had a “subscale” for how vision/eye problems affected the participants physical productivity, i.e. their ability to walk, lift things, turn their bodies. In other words, the subscale dealt with productivity loss from inhibited bodily abilities because of vision problems. Vision problems are not likely to affect people’s ability to walk or perform basic physical motor functions (at least not the type of vision problems referred to in this paper), but the subscale was included in spite of that to be able to calculate the total percentage of productivity loss. There is a risk that the questions within this
subscale were confusing to the participants, but apparently not too deterrent as all participants answered all questions in the WLQ.
36
he approved the translation after one or two minor adjustments. The risk of mistranslating any of the questions is limited by the simple wording of the questionnaire, see appendix 8.
Future research
It would be interesting to do larger studies that research the prevalence of CVS and CI among office workers and further investigate the association between the vision
problems and productivity loss. Preferably where more possible confounding factors are more thoroughly researched, such as burnout, which could have an impact on the
productivity loss. It would also be interesting to research if there are any differences in the work intensity for office workers with and without confirmed CI and measure the number of micropauses (when the computer worker need to look away from the screen) during reading a text to get another measurement for productivity loss. Later on, it would also be interesting to do intervention studies to research if it is possible to increase the visual health and reduce the productivity loss among office workers. It would also be interesting to investigate other symptoms related to office work and its associations with productivity loss, such as neck- and shoulder pain or burnout.
Conclusion
