Validation of WHOQOL-BREF in Malayalam and Determinants of Quality of Life Among People With Type 2 Diabetes in Kerala, India

Abstract

Quality of life (QOL) is an important health outcome in people with chronic conditions like diabetes and WHOQOL-BREF is a popular instrument used worldwide to assess QOL. However, QOL varies considerably from society to society depending on the culture of the person. Hence, the WHOQOL-BREF was translated to the local language, Malayalam. This article attempts to establish reliability, construct and discriminant validity of the translated WHOQOL-BREF, and determinants of QOL among people with type 2 diabetes. A cross-sectional study was undertaken among 200 patients with diabetes attending a primary care center in a rural area of Kerala, India. The translated version of WHOQOL-BREF was found to be internally consistent (Cronbach’s α = .86) and demonstrated discriminant and construct validity. Education was found to be an independent determinant of QOL in the physical, psychological, and environmental domains. Thus, the translated version had good psychometric properties and education was an independent determinant of QOL in 3 of 4 domains.

Quality of life (QOL) is an important health outcome¹ and considerations of QOL are gaining increasing importance in evaluation of health policy and medical intervention.² Currently, measures of disease alone are insufficient to measure health status and subjective measures of health and well-being provide a multidimensional view point.³ However, QOL varies considerably from society to society depending on the culture of the person.^1,4 World Health Organisation(WHO) defines quality of life as “an individual’s perception of their position in life in the context of the culture and value systems in which they live in relation to their goals, expectations, standards and concerns.”⁵ The World Health Organization’s WHOQOL-BREF is an instrument that has been widely used internationally, validated and is generalisable to many populations.⁵ Health-related QOL also signifies the impact of chronic conditions like diabetes⁶ and has been found to be far worse among those with diabetes.⁷

There are about 75 million people living with diabetes in the Southeast Asian region of which 66 million are in India with a national prevalence of 8.3%.⁸ The sheer numbers of cases of diabetes mellitus point to a significant burden⁹ and the resultant complications too affect the quality of life.^10–12 People with diabetes often feel challenged by their disease and its day-to-day management demands.¹ Kerala, a southern state in India where the study was carried out is in an advanced state of epidemiological transition and leads India in terms of the number of people with diabetes.^13,14 Perception of the patients about the illness and how it affects them are moderated by ethnicity, culture, beliefs, and social realities. Although WHOQOL-BREF exists in 30 languages,⁵ each state in India has its own sociocultural identity, and there are not many instruments that measure QOL in Malayalam, the language spoken in Kerala. WHOQOL-BREF was translated in accordance with the WHO policy and specifications and the process, challenges, and content validity have been mentioned in a previous publication.¹⁵ Here, we attempt to determine reliability, construct validity, and discriminant validity of the translated WHOQOL-BREF among people with type 2 diabetes and assess determinants of QOL.

Methods

This was a cross-sectional study conducted after approval by the institutional ethics committee. The WHOQOL-BREF instrument with 26 items—across 4 domains physical (7items), psychological (6 items), social relationship (3 items) and environment (8 items) was used. There were 2 global scores of overall QOL (1 item) and overall satisfaction with health (1 item). All the 200 patients with type 2 diabetes accessing the services of a primary care center from January to March 2013 were interviewed with the first part of the questionnaire, which dealt with sociodemographic information, diabetes-related information, and with the second part, which was the translated WHOQOL-BREF, after getting written informed consent. Inclusion criteria were persons with type 2 diabetes for more than 6 months. People with chronic and painful health conditions like cancer, spine injury, and psychiatric morbidity were excluded from the study. Discriminant validity was measured among a subset of 30 by administering Social Desirability Scale 17 (SDS 17) and Digit Forward Task Test (DFT). Convergent validity was measured with PGI Well-being Scale, which assesses well-being administered to another subset of 30. The DFT measures attention and the social desirability scale assessed the perceived social desirability of their behaviors. Internal consistency was estimated using Cronbach’s alpha and split half reliability. Correlation and independent-sample t test were used to determine discriminant and convergent validity. Exploratory factor analysis was also done. A p value of <.05 was considered significant. Quality of life scores above 75th percentile were considered as good QOL and determinants analyzed.

Results

The mean age of the sample was 54.96 ± 10.61 years and other characteristics are as in Table 1. The overall QOL and health-related QOL varied from 60 to 80 (interquartile range [IQR]) with a median of 80 and from 40 to 80 (IQR) and a median of 60, respectively, indicating a fair QOL.

Table 1.

Sociodemographic Distribution of the Study Population.

Characteristic	n (%)
Age distribution, years, n = 200
<30	3 (1.5)
31–70	184 (92.0)
>71	13 (6.5)
Gender, n = 200
Male	76 (38.6)
Female	124 (61.4)
Occupation, n = 200
Homemaker	89 (44.5)
Unemployed	30 (15.0)
Employed	81 (40.5)
Education, years of schooling, n = 199
>12	29 (14.5)
8–12	85 (42.7)
<8	85 (42.7)
Type of family, n = 198
Nuclear	108 (54.5)
Joint	90 (45.4)
Religion, n = 183
Hindu	130 (71.0)
Christian and Muslim	53 (28.94)
Marital status, n = 199
Ever married	194 (97.4)
Never married	5 (2.5)
Comorbidities, n = 183
No	63 (34.4)
Yes	120 (65.5)
Duration of diabetes,years n = 194
0.5 to 5	78 (40.2)
5–10	61 (31.4)
>10	55 (28.3)

But domain-wise IQR range varied from 42.85 to 64.28 (median 53.57), from 45.83 to 70.83 (median 58.33), from 50 to 75 (median 66.6), and from 53.12 to 71.87 (median 62.5), in the physical, psychological, social, and environmental domains, respectively. Among the domains, physical domain fared the worst with the least median score and the social and environmental were relatively better, perhaps due to better social relationships, better physical security, and better health care access. At a more subjective level, the questionnaire appeared to touch a cord with people eliciting emotions like sadness, frustration, and so on.

Validation

Reliability Analysis

Reliability analysis showed a high overall Cronbach’s alpha of .86 (p < .001). The internal consistency in the physical domain was at 0.734, psychological domain at 0.652, social domain at 0.615, and environmental domain at 0.726. The tool was also found to be reliable by the Guttmann’s split half technique at 0.813.

Discriminant Validity

As anticipated, there was no significant association between the 2 measures, DFT and SDS 17 and QOL domain scores indicating discriminant validity. However, SDS 17 score showed an association with the psychological domain of WHOQOL-BREF (0.373, p < .05), which could be attributed to the psychological dimension of acceptance (Table 2).

Table 2.

Convergent and Discriminant Validity.

Discriminant Validity			Convergent Validity
	Digit Forward		Social Desirability		PGI Well-Being
Domain	Mean Difference	P	Correlation Coefficient	P	Correlation Coefficient	P
Physical	−4.9	.23	0.21	.22	0.45	.007
Psychological	−6.15	.20	0.37	.03	0.44	.01
Social	0.83	.87	−0.10	.55	0.08	.64
Environmental	−4.09	.40	0.14	.41	0.11	.51

Convergent Validity

The physical and psychological domains of the translated WHOQOL BREF showed a significant correlation with PGI Well-Being Scale at 0.459 and 0.440, respectively (p < .007 and p < .01), thus establishing convergent validity (Table 2).

Construct Validity

Analysis showed that all cross domain correlations were lower than item domain correlation (data not given here). However, 2 items—“Satisfaction with oneself” (psychological domain) and “Ability to move around the home” (physical domain)—had strong correlation with physical (0.5) and psychological (0.47) domains, respectively, probably due to overlap and inadequate conceptualization. Another poorly understood item was about “body image,” which poorly correlated with psychological domain at 0.3, and “leisure time entertainment” in the environmental domain was poorly understood with a correlation of 0.2. QOL is a complex construct and in order to establish its construct validity WHOQOL-BREF domain scores were compared with general single item scores, indicating global QOL and global health-related QOL (Table 3). Overall QOL was strongly associated with physical, psychological, social, and environmental domains at p < .001. Health-related QOL was highly significantly associated with physical and psychological domains (p < .00) and also with social and environmental domains at p < .05.

Table 3.

Association of Domains and General Facet Items of WHOQOL-BREF.

	Overall Quality of Life		Overall Health
Domain	Correlation Coefficient	P	Correlation Coefficient	P
Physical	0.48	.00	0.47	.00
Psychological	0.47	.00	0.44	.00
Social	0.28	.00	0.16	.02
Environmental	0.39	.00	0.16	.02

Exploratory Factor Analysis

WHOQOL-BREF with 26 items was subjected to exploratory factor analysis on the basis of response of 200 respondents. Exploratory factor analysis using the principal components factor extraction method with a nonorthogonal rotation (varimax) was carried out to frame the domains. Following the rotation, 4 domains as in the original WHOQOL BREF were obtained. Domain 1, which in the original was the physical domain, retained 4 of the original 7 items with highest correlation at 0.8 for the “working capacity” followed by 0.78 for the “activities of daily living,” 0.5 for “the ability to move around,” and 0.44 for the “dependence on medication support.” On the third domain derived as a result of rotation which corresponds to environmental domain in the original questionnaire, out of the original 8 items, 4 items showed high levels of correlation, namely “sense of safety and security” with a correlation of 0.7, “opportunities for acquiring information” with a correlation of 0.6, “healthy surroundings” with a coefficient of 0.53, and “leisure opportunities” with a coefficient of 0.4. In the other 2 domains derived out of the rotation, significant overlap was found between the items of original physical and psychological domains.

Determinants of Quality of Life

The determinants of QOL in this sample were examined. In the physical, psychological, social, and environmental domains the percentage of people occupying more than 75th percentile corresponded to 7.1%, 16.2%, 17.25%, and 40.1%, respectively, indicating an overall poor QOL in all domains. The respondents experienced a relatively better QOL only in the environmental domain.

The domains of WHOQOL-BREF were analyzed for determinants like gender, education, occupation, religion, type of family, marital status, comorbidities, and duration of diabetes. In the physical domain of QOL with increasing education a highly significant increase in good QOL scores was observed (p < .001). Respondents belonging to nuclear families (11.4%) were found to be significantly associated with better QOL scores compared with 2.2% in joint families (p < .05).Surprisingly, 40% of unmarried people seemed to enjoy better QOL in comparison with 6.3% of married individuals (p < .05). About 13.1% with no comorbidities had high QOL scores compared with 4.1%with comorbidities (p < .05). There was no significant association with other variables.

The psychological domain also demonstrated a highly significant association with education, with 40.7% of those with a more than 12 years education having higher QOL compared with 23.6% of those with 8 to 12 years of education (p < .001). The social domain of QOL was also significantly related to education though the increase in percentage of QOL scores was marginal from 35.29% to 43.9% to 44.8% from less than 8 years of schooling to more than 12 years of schooling (p < .05). The other factors were not significantly associated.

The environmental domain also showed a significant increase in QOL scores with education from 11.7% in the less than 8 years schooling group to 41.3% in the more than 12 years schooling group (p < .003). In this domain, an association of QOL with gender was observed with 24% of men having good QOL scores compared with13.1% of women (p < .05).

Logistic regression carried out among the significant variables showed that education was an independent determinant of QOL across physical (>12 years of education, odds ratio [OR] =7.22, 95% CI = 2.48–18.35, p < .001; psychological (8–12 years of education, OR = 3.3, 95% CI = 1.5–7.22; >12 years of education, OR = 5.4, 95% CI = 2.08–14.39, p < .0010); and environmental domains (8–12 years of education, OR = 4.8, 95% CI = 2.16–10.63, p < .001; >12 years of education, OR = 5.29, 95% CI, 1.96–14.25; p < .001) (Table 4).

Table 4.

Education as an Independent Determinant of Quality of Life.

	Score (%)(>75th percentile)	Total	Odds Ratio	95% CI	P
Physical domain
Education, years of schooling
<8	20 (23.5)	85	1
8–12	28 (34.1)	82	1.68	0.856–3.319	.131
>12	20 (68.9)	29	7.22	2.48–18.35	<.001
Psychological domain
Education, years of schooling
<8	4 (5.1)	78	1
8–12	17 (23.6)	72	3.3	1.5–7.22	.003
>12	11 (40.7)	27	5.46	2.08–14.39	.001
Environmental domain
Education, years of schooling
<8	10 (11.7)	85	1
8–12	32 (39.0)	82	4.8	2.16–10.63	<.001
>12	12 (41.3)	29	5.29	1.96–14.25	.001

Discussion

Validation

The translated WHOQOL-BREF was found to have good internal consistency. The tool was also found to be reliable by Guttman split half technique at .813.The tool was found to demonstrate good discriminant, good convergent validity, and fair construct validity. Previous studies have also validated the tool internationally³ and in Indian settings,¹⁶ among adolescents.¹⁷

The factors derived as a result of exploratory factor analysis on physical and environmental domains were in consensus with the items and domains in the original version. However overlap of the items between the physical and psychological domains was also observed. Cross-domain correlation also showed an overlap between physical and psychological domains and 2 items of “satisfaction with oneself” and “ability to move around home” had an equally strong correlation with both domains. Similar overlap between physical and psychological domain has been reported by Skevington et al³ from the Indian site, Madras. This can probably be improved by changing the wording to reinforce the intended concept.³ Two items in the item total domain correlation also showed a weak correlation which needs further studies. This also points to the moderating effect of cultural perception on the concept of QOL across different domains of QOL. The psychometric properties of translated WHOQOL-BREF are good, though it is limited by the fact that confirmatory factor analyses could not be done due to insufficient sample size.

Determinants of Quality of Life

The percentage of people with good quality of life was low and varied from 7% to 17% in the physical, psychological, and social domains, which is supported by other studies.^18,19

Education was found to be an independent determinant of good QOL in all the domains except social. Education as a predictor of QOL has been established across different populations.^20–22 Education reduces distress largely by way of paid work, nonalienated work, and economic resources, which are associated with high personal control.^22,23 Indian studies also report increase in QOL with increasing education.^19,24

Conclusion

The translated WHOQOL-BREF is a psychometrically sound tool with high internal consistency and good discriminant and convergent validity and can be used in the Malayalam-speaking population. Construct validity can be further established by carrying out studies in larger populations and by changing the wording and semantics of the four items. Education is found to be an independent predictor of good QOL in 3 of the 4 domains of WHOQOL-BREF indicating the critical role that education has to play in influencing health care outcomes in chronic diseases like diabetes.