The Measurement of Disability-Related Stress in Wheelchair Users

Phase 1: development of the instrument 

The common areas of disability-related stress were established through the development of an initial survey in which participants provided written information regarding the top 5 most stressful aspects of having a physical disability. The participants were asked to rank their written statements from 1 to 5 in order of most stressful to least stressful. This information was gathered from the first 25 people to respond to advertisements. The respondents were 13 men and 12 women (age range, 25−72y), with SCI or PPS being the predominant disabilities. These 25 respondents were invited to participate in phase 2 of the study; however, because the questionnaires were anonymous, it is unknown how many of the initial 25 respondents are included in the final sample of 119.

Once the initial surveys were returned, the 2 chief researchers independently read the written responses and generated a list of the common areas of disability stress. The individual lists were then compared and areas of agreement and disagreement examined. This resulted in the generation of a list of the most commonly reported areas of disability-related stress. In particular, time was spent on the wording of domains such that they most accurately captured the information provided.

The written responses from participants were reread by the researchers and 2 volunteer raters from nondisability-based backgrounds. Responses were categorized into the nominated domains. Domains with less than 10 responses were discarded, leaving only those domains that have been mentioned most frequently and thus by most people. The final 5 domains were as follows: (1) access, (2) perceptions and attitudes of others, (3) social and sexual relationships, (4) physical health, and (5) adjustment and loss of independence. Specific items were written for each of these final domains and, when possible, the wording of items was based on specific statements made by participants in their written responses. A colleague with a disability then provided independent feedback to ensure the wording of the items was clear and nonoffensive. Forty items were generated with 7 to 9 items under each domain. The items were then randomized. This process of instrument development is based on recommendations by Jackson and Furnham.33

In completing the questionnaire, participants were asked to think about the disability-specific situations mentioned in the items and to indicate how stressful these situations are for them on a scale of 1 to 5 (1, not at all; 2, slightly; 3, moderately; 4, considerably; 5, highly). Participants were also given the option of circling NA (not applicable) for situations they have not experienced. Many investigations32 in psychologic assessment have focused on rating scales with adjectival descriptions. The main advantage of this format is that it makes sense to the respondent and can therefore be easily completed. Jackson and Furnham33 recommend that the minimum number of categories used by raters should be in the region of 5 to 7 because reliability coefficients drop as fewer34 or greater35 numbers of categories are used. The final instrument was called the PDSS.

Phase 2: administration of the instruments 

The PDSS and other measures were sent via mail to participants. Other measures included the General Health Questionnaire−28 (GHQ-28)36 and the WHO Quality of Life (WHOQOL-BREF),37 Australian version. The GHQ-28 is a 28-item, self-administered screening test that assesses severe depression, anxiety and insomnia, social dysfunction, and somatic symptoms. The GHQ-28 can be used to establish presence of psychiatric illness in a population. By using the GHQ total score, participants with a score of 23 or below are classified in the nonpsychiatric group, whereas those with a score of 24 or above are in the psychiatric group.38 The GHQ-28 has been shown to be a valid and reliable measure in many populations including spinally injured outpatients.39

The WHOQOL-BREF was developed from the larger WHOQOL-100 datasets. The shorter version was preferred because respondents may have had physical difficulty completing the longer version. The Australian WHOQOL-BREF contains 26 items that are rated on a 5-point scale. There are 4 domains: physical, psychological, social relationships, and environment domains, and all show good internal consistency and test-retest reliability.40 For the purposes of this study, this instrument will be referred to as the QOL measure.

Principal components analysis 

Before analysis, an inspection of the item correlation matrix revealed a number of significant coefficients. A Kaiser-Meyer-Olkin value of .82 indicated that partial correlations were small. Bartlett’s test of sphericity, sensitive to zero-order correlations in the correlation matrix, reached statistical significance (χ7032=2311.04, P<.001). This indicated factorability of the data.

Because it was expected that the factors would be related, a PCA with an oblique rotation (obliminal rotation with Kaiser normalization) was performed. The PCA revealed 9 factors; the first 4 of these closely resembled the factors that had been created in the initial instrument development following phase 1 responses. The last 5 factors consisted of only 2 items in each factor and were considered uninterpretable.

A second PCA was performed to force the extraction of the 5 a priori factors. An oblique rotation was performed (obliminal rotation with Kaiser normalization) to account for the expected correlations between the factors. The first 4 of the resulting factors showed a strong resemblance to the original factors; however, the fifth factor consisted of only 2 items with eigenvalues greater than .4. These items had low internal consistency (α=.18) and were considered uninterpretable. The fifth factor was thus deleted. The remaining 4 factors showed significant correlations (r range, .308−.575; P<.01).

Analysis of the 4 factors indicated that a further 14 items met criteria for exclusion from the item set, thus leaving 22 items. Items were excluded if they showed (1) eigenvalues less than .4 or (2) significant loading on more than 1 factor.26 A total of 18 items were removed from the original 40 items. Deleted items can be obtained from the authors by request. Table 1 shows the factor loadings of the remaining items on their factors. Comrey and Lee46 suggested that factor loadings in excess of .71 are considered excellent, .63 very good, .55 good, .45 fair, and .32 poor. The first factor, labeled access, accounted for 33.7% of the variance and represented the aspect of disability-related stress that is associated with limited access and reduced motility (independent and spontaneous movement) in and out of environmental structures. The second factor, labeled physical, accounted for 8.4% of the variance and refers to the stress of reduced bodily movements and functional ability. On a broader level, this factor also refers to the stress of reduced involvement in physical activities and to the decreased ability to individually contribute to group tasks. The third factor, labeled social, accounted for 7.9% of the variance and covers the area of stress that is experienced through social stigma and social interactions with other members of society. This factor also encompasses the stress associated with sexual relationships, friendships, and changes to social activities. Finally, the fourth factor, labeled burden of care, accounted for 7.2% of the variance and relates to the stress and burden of having and managing numerous health care needs and of relying on others to help in health care areas (see appendix 1 for the PDSS and scoring instructions).

Table 1. Factor Loadings for the PDSS Items (N=119)

		Factor 1	Factor 2	Factor 3	Factor 4
	Items	Access	Physical	Social	Burden of Care
	1. When you have been unable to travel independently due to inaccessible places	.83
	3. When you have been carried up or down stairs	.61
	21. When you have arrived at a place that has stairs, but has no ramps or elevators	.70
	40. When you have been told a place is accessible, but find it isn’t accessible when there	.75
	13. When you have not been able to do some activities that you used to enjoy		.58
	24. When you have been unable to physically help with manual jobs		.75
	27. When you have been unable to physically move the way you used to move		.57
	32. When you felt there was nothing you could do to change things		.75
	35. When you have been unable to physically help others when they need it		.67
	2. When you have not been invited to social activities as much as you used to			.71
	8. When private issues have been made public			.43
	9. When you have used a taxi service			.55
	10. When your disability has affected your relationships and friendships			.84
	15. When others have not recognized your sexual identity or sexual desires			.62
	19. When your disability affected the development of intimate/sexual relationships			.74
	25. When others have seen you as a person in a wheelchair before they have seen the person you are			.71
	7. When people have tried to help but have made things more difficult				.43
	11. When you have had to deal with carers or helpers				.63
	18. When you have had to rely on others for help				.43
	20. When you have had to manage a number of health care needs				.88
	29. When your health care needs have required time and energy				.83
	33. When others have not considered your needs or abilities				.45

Internal consistencies for the 4 factors were within acceptable ranges (α range, .78−.83), indicating homogeneity of the scale. Table 2 outlines the properties and internal consistencies of the PDSS, including floor and ceiling statistics.

Table 2. Internal Consistencies and Other Properties of the PDSS (N=119)

	PDSS Factors	No. of Items	Range of Scores		Mean ± SD	Cronbach α
			Minimum (% min)	Maximum (% max)
	Access	4	0(0)	20(7.6)	14.95±4.1	.79
	Physical	5	0(0)	25(9.2)	18.51±4.8	.82
	Social	7	0(0)	35(0)	21.89±6.8	.83
	Burden of care	6	0(0)	30(0.8)	18.80±5.0	.78

The relationship between the PDSS and GHQ 

Table 3 summarizes the findings from 5 multiple regression analyses performed by using PDSS factors entered simultaneously as predictors of the 4 GHQ subscales and the GHQ total score. As reported in table 3, 7% to 23% of the GHQ subscales and 17% of the GHQ total score were predicted by scores on the PDSS factors. The burden of care factor was the best predictor of variance on the GHQ total score and on all the GHQ subscales other than the social dysfunction subscale (which was best predicted by physical). Some of the variance in GHQ severe depression was additionally accounted for by the access and physical factors.

Table 3. PDSS Factors as Predictors of the GHQ Subscale (N=119)

		Somatic Symptoms		Anxiety and Insomnia		Social Dysfunction		Severe Depression		Total Score
	Measure	R2	β	R2	β	R2	β	R2	β	R2	β
	PDSS	.07		.15‡		.10†		.23‡		.17‡
	Access		.06		.00		−.05		−.23⁎		−.08
	Physical		−.11		.04		.22⁎		.28†		.14
	Social		−.01		.07		−.19		.11		.02
	Burden of care		.25⁎		.32†		.24		.27⁎		.35†

⁎ P<.05. † P<.01. ‡ P<.001.

Classification of participant’s scores on the GHQ into psychiatric or nonpsychiatric “caseness” groups38 allowed for a comparison across PDSS factors. MANOVA revealed an overall psychiatric group effect (Hotelling T2=.14, F4,114=3.98, P<.01). Follow-up univariate analyses indicated that those in the psychiatric group showed significantly higher stress levels in areas of physical, social, and burden of care, whereas those in the nonpsychiatric group showed lower stress scores on these factors (table 4). There were no differences between psychiatric and nonpsychiatric groups on the access factor.

Table 4. Relation Between PDSS Factors and the Psychiatric and Nonpsychiatric Groups on the GHQ-28 (N=119)

	PDSS Factors	Group (GHQ-28)	Mean ± SD	F
	Access	Nonpsychiatric	13.44±4.33	1.66
		Psychiatric	14.49±4.52
	Physical	Nonpsychiatric	16.75±5.14	4.80⁎
		Psychiatric	18.83±5.02
	Social	Nonpsychiatric	17.29±7.78	5.04⁎
		Psychiatric	20.42±7.06
	Burden of care	Nonpsychiatric	16.13±5.17	13.93†
		Psychiatric	19.61±4.77

NOTE. Degrees of freedom were 1,116 for all analyses of variance.

⁎ P<.05. † P<.001.

The relationship of the PDSS to QOL 

Table 5 summarizes the findings from 4 multiple regression analyses performed by using PDSS factors entered simultaneously as predictors of the 4 QOL subscales. As shown, 12% to 31% of the QOL subscales were predicted by scores on the PDSS factors. Specifically, physical accounted for the most variance in QOL physical health, whereas social was the best predictor of QOL social relations. The variance in QOL psychological health was best predicted by burden of care, physical, and access factors. Finally, 12% of the variance in the QOL environment subscale was predicted by the PDSS factors, with a trend for the burden of care factor to be the best predictor.

Table 5. PDSS Factors as Predictors of the QOL Subscales (N=119)

	Measure	QOL Physical Health		QOL Psychological Health		QOL Social Relationships		QOL Environment
		R2	β	R2	β	R2	β	R2	β
	PDSS	.13†		.22‡		.31‡		.12†
	Access		.01		.23⁎		.26		.05
	Physical		−.28†		−.31†		−.15		−.16
	Social		.10		.07		−.64‡		−.06
	Burden of care		−.19		−.35†		.20		−.22§

⁎ P<.05. † P<.01. ‡ P<.001. § P<.10 (for trend).