Development and validation of a questionnaire assessing the quality of life impact of Colour Blindness (CBQoL)

This study was an online survey analysed using standard questionnaire validation methods, such as principal components analysis. People with CB were involved in the design and implementation of this study.

This first part of this paper describes the development of the questionnaire, and the second part describes the validation of the questionnaire. In the second part, the scores of people with CB are compared with the scores of people with normal colour vision.

The setting of this study was online.

Participants were recruited from various sources, mostly online (details in the section Initial Validation of Final CB Questionnaire, below). They were categorised as being CB or not, as described in the Materials section below. The authors use the common term ‘colour blindness’ (CB) for the sake of clarity, but recognize that participants in fact had varying degrees of colour vision deficiency (CVD). Characteristics of the sample are shown in Table 1 below.

Ethical approval was granted by the University College London Research Ethics Committee (Project ID: 4075/004). Informed consent was given by the participants before filling in the questionnaire.

Table 1 shows the characteristics of the sample.

As expected, there were significantly more men than women with colour blindness (χ ² = 118.98, df = 2, p = 1.5 × 10⁻²⁶). Of the 65 CB participants, four (6%) reported that their CB was acquired rather than congenital. There was no significant difference in the socioeconomic background distributions in the three vision groups, with 56–68% coming from a professional background, 23–44% from a manual background, and 0–9% from an intermediate background.

The mean (SD) age of the CB group (44.8 ± 15.6 years) was significantly older than that of the NV group (30.1 ± 15.1 years) (p = 9 × 10⁻¹¹). In order to reduce any effect of age on outcomes, matching of participants in each group by age was carried out.

Of the 56 CB participants aged 18 to 65 years old, it was possible to match 30 of them by age to the nearest year to one or more NV community controls. This age limit of 65 years was chosen both because it spans the average working age in the UK and avoids the general declining of vision which is common with older age. Where there was more than one person in a group of the same age, the mean of their scores was used. For example, for participants aged 30 years old, there was one CB and three NV controls, thus the sole CB score was compared to the mean of the three NV scores. In some cases there was more than one CB participant of the same age e.g. at age 45 the mean of four CB participants was paired with the mean of six controls. Using this process, 30 age-matched pairs were possible from 46 CB and 199 control participants. Because men and women typically have different rates and types of colour blindness, the groups were further subdivided by sex. Fewer age-matched pairings were possible within each sex: 11 pairings of men (16 CB and 18 controls) and five pairings of women (five CB and 20 controls).

A further step in validating new questionnaires is testing how much they are in agreement with existing validated questionnaires measuring similar constructs. This is known as concurrent validity, and acceptable concurrent validity is indicated by a Pearson’s correlation coefficient (r) of 0.5 or more [20]. The criterion by which the new questionnaire subscales were measured was the Ishihara Colour Blindness Test [15], which assesses colour vision acuity. Thus concurrent validity between the CBQoL questionnaire and the Ishihara was assessed in this way.

To assess concurrent validity with health and psychological aspects of QoL, two other questionnaires were used. Firstly, the QoL mental health subscale of the SF-36 [17], on which lower scores indicate worse mental health. Secondly, the Positive Mindset Index (PMI) [18] was used, on which higher scores represent a more positive state of mind.

Responses were on a 6-point Likert scale from 1 = A severe problem, to 6 = No problem, with an option for ‘not applicable’. The scores on the subscales (Table 2) were converted to means (Table 3) with a maximum score of 6 and minimum of 1. The 36 items relating to problems caused by colour confusion were: career limitations, avoiding aspects of work, underachievement, problems with coloured charts, coloured text, buying clothes, driving, reading maps, avoiding discussions involving colours, caused problems in family, social activities, socializing, ripeness of fruit, food cooked fully, choosing groceries, coloured medication, sunburn, mole on skin, urine, blood in faeces, avoiding activities, felt let others down, embarrassment, depression, anxiety, unconfident, feeling different, self-esteem, annoyed, worried, problems with sports, problems in dim lighting, pain, concentration, confusion, and lethargy. For the purposes of this initial research into CVD, there were also free text questions to allow participants to state other emotional and physical symptoms not in the list; the free text responses did not add a great deal to the statistical information and are not discussed below, and are not part of the CBQoL.

Table 3 shows that the CB participants scored significantly lower on the Ishihara and the three CBQoL subscales, but not significantly lower on the mental health subscale of the SF-36 HRQoL or PMI.

For the CB participants, there was a positive correlation between the Ishihara test and the CBQoL subscales: Health & Lifestyle: r = .533, n = 91, p = 5 × 10⁻⁸; Work: r = .599, n = 91, p = 5 × 10⁻⁸; Emotions: r = .315, n = 91, p = 0.002. This shows acceptable concurrent validity of the CBQoL for the Health & Lifestyle and Work subscales, though weaker evidence for the concurrent validity of the Emotions subscale. In contrast, the correlations between the Ishihara test and QoL subscales in the control group were weak (average r = .106).

In a further validation of the Health & Lifestyle and Emotions subscales, for the CB group only there was a statistically significant, moderate positive correlation between the CBQoL Health & Lifestyle subscale and the mental health subscale of the SF-36 (r = .297, n = 63, p = 0.018), and there was a statistically significant, moderate positive correlation between the CBQoL Emotions subscale and the PMI (r = .393, n = 63, p = 0.001). Although these correlations are weaker than ideal for concurrent validity, they are stronger than those for the control group (CBQoL Health & Lifestyle subscale and the SF-36: r = .200, n = 322, p = 0.0003; CBQoL Emotions subscale and the PMI; r = .057, n = 321, p = 0.308), which supports ‘known groups’ validity.

In addition to its primary aim of developing the first QoL measure specific to CB, the present study provides a quantitative estimate of the impact of this common but often overlooked condition. This study is the first to assess the QoL impact of problems with colour vision in people with colour blindness compared to normally sighted controls. The three subscales show good psychometric properties in terms of factor loadings, Cronbach’s alpha and other relevant measures.

The CBQoL has further potentially useful applications e.g. in contrasting the effects of congenital vs acquired colour blindness, the effects of different types of congenital colour blindness (e.g. across protan, deutan, tritan axes) and the varying impact of different acquired diseases (e.g. retinal pathologies vs optic neuropathies). A future study of an intervention for people with acquired colour vision problems might use the CBQoL to measure changes in QoL from pre- to post-treatment.

The cause of CB was not explored in this study. It may be that some forms of CB have a much less - or much greater - impact on QoL than we observed here. Future work will explore the variation in impact between congenital and acquired forms of CB.

The method of recruitment to this study may have led to sample bias. Most participants were recruited online, which limits the sample to those who had access to the internet and/or involved in colour blindness forums, or recruited through professional groups. This presumably contributed to a relatively high proportion of participants in each group being from a professional background, though this was not significantly different across the CB and NV groups. It is worth noting however that CB people may have career limitations and different employment due to colour difficulties than NV people [7]; inadvertent sampling bias towards professional groups may therefore exclude people whose careers have been most negatively affected by CB and therefore underestimate the impact of CB on work related QoL. Future research using the CBQoL might benefit from recruiting from non-online sources (e.g. in clinics) and recruiting more working class participants.

It is noted that the Ishihara test was not administered face-to-face by the research team but was administered online in uncontrolled conditions i.e. in participants’ own homes, where backlighting and monitor settings were not standardised by the experimenters; we cannot therefore externally validate the accuracy of the Ishihara scores. However the fact that the Ishihara scores were strongly correlated with the Work and Health & Lifestyle CBQoL subscales indicates some evidence of good concurrent validity.