An exploratory factor analysis of Telugu versions of the WHO well-being index, the UCLA-loneliness scale and the perceived criticism scale in caregivers of schizophrenia patients

ABSTRACT

Background:

The lack of locally validated instruments hinders proper evaluation and management of caregiver burden.

Aim:

To perform an exploratory factor analysis (EFA) of Telugu-translated scales assessing psychological factors related to caregiver burden.

Methods:

This cross-sectional study was conducted at three sites: Among 100 caregivers of schizophrenia patients using Telugu-translated versions of the WHO Well-being Index (WHO WBI), the UCLA Loneliness Scale (ULS), and the Perceived Criticism Scale (PCS). EFA was used to identify the factor structure.

Results:

Principal component analysis revealed that the items of the scales- WHO-WBI, ULS, and PCS were significantly loaded on one factor each, with eigenvalues of 3.810, 2.504, and 2.105, respectively, and the substantive factors represent “mood”, “loneliness”, and “criticism”, respectively. Cronbach’s α coefficient was 0.9 for WHO-WBI and ULS, and 0.78 for PCS, indicating an acceptable level of internal consistency.

Conclusion:

Telugu versions of WHO-WBI, ULS, and PCS hold fair-to-good structural validity and internal consistency.

INTRODUCTION

Schizophrenia has an early onset and continuous course with frequent relapses, thereby causing significant dysfunction in multiple aspects of an individual’s life.[1] Recent studies from India reveal that 90% of patients live with their families.[2] Addressing the negative psychosocial impact of caregiving is of utmost importance, as it also impacts patient outcomes. Many studies have evaluated the impact of burden on the caregivers of patients with schizophrenia.[3] Some of the negative caregiving outcomes include loneliness, psychological morbidity, poor well-being, and expressed emotions.

Most scales that assess the psychosocial outcomes are self-rated and in English, which limits their applicability in India due to language and cultural differences.[4] Hence, translation, adaptation, and validation are necessary for self-rated tools. Majority of previous Indian studies have been conducted primarily in northern states. Moreover, limited studies from southern states have been restricted to Kerala, Tamil Nadu, and Karnataka.[3]

Scales such as the WHO well-being index (WHO-WBI),[5] UCLA loneliness scale (ULS),[6] and perceived criticism scale (PCS),[7] which are commonly used to measure social connectiveness, well-being, and expressed emotions, have good validity and reliability. However, these self-rated tools are not available in Telugu, thereby restricting their use in the Telugu-speaking states. Hence, we aimed to translate and validate the Telugu version of three scales: the WHO-WBI, ULS, and PCS.

METHODOLOGY

This was a cross-sectional study conducted over a period of one-year (August 2024–July 2025). The study was done as part of a multicenter study, “Correlates of caregiver burden in schizophrenia”. The study involved Telugu translation using the standard guidelines[4] and exploring the structural validity of the WHO-WBI, ULS, and PCS.

Scales

WHO-WBI, is a 5-item questionnaire used to measure psychological well-being.[5] Each item is scored on a 6-point Likert scale. The ULS, with 3-items, is brief and feasible for use in large-scale studies. It measures loneliness and alienation.[6] Each item is rated on a scale of 1 to 3. PCS, a 3-item questionnaire, measures expressed emotion.[7] Each item is scored on a scale of 1 to 10.

Translation

The scales were translated using standard guidelines[4] into Telugu by two translation experts, fluent in both languages. The best-translated version was back-translated into English by two mental health professionals. Discrepancies were resolved through discussions with experts and the principal investigator. Serial checks of the original and the back-translated version were performed till linguistic equivalence was agreed. To further validate, the final translated version was given to 10 bilingual caregivers at two time points (a week apart), and the inter version correlation coefficient using Haccoun’s technique was assessed. The final Telugu version was then used for EFA.

Data collection

The study was conducted in the outpatient and inpatient settings of the psychiatry department at three centers in two Telugu-speaking states.Approval from the Institute’s Ethics Committee of all centers was obtained, and participants were enrolled after obtaining informed consent.

The study included caregivers of patients with schizophrenia. Caregivers were aged over 18 years, of any gender, and had been caring for patients for at least past 1 year prior to recruitment. All caregivers were able to read, write, and speak Telugu. Caregivers with any form of cognitive impairment, organic brain syndrome, intellectual disability, any psychiatric disorder, or substance use disorder (except nicotine) were excluded. Caregivers with patients who had comorbid organic brain syndrome or intellectual disability were also excluded. The collected data included the patients’ and caregivers’ socio-demographic profiles. The caregivers rated the Telugu-translated versions of the WHO WBI, ULS, and PCS. A uniform standard operating procedure for data collection was followed at all centers. One designated investigator from each center collected the completed forms that were filled privately to minimize the social desirability and acquiescence bias.

Statistical analysis

Sample size estimation

The sample size was determined based on the maximum number of items across the three scales (5 for WHO-WBI). Using the “20 participants per item” rule-of-thumb for instrument validation,[8] the sample size estimate was 100 participants. This was achieved through convenience sampling.

Exploratory factor analysis

The EFA was performed to analyse the factor structure of the scales: WHO WBI, ULS, and PCS. The analysis was conducted using IBM SPSS Statistics version 20 software (IBM Corporation, New York, USA). Assumptions for the EFA were analyzed using the Kaiser–Meyer–Olkin (KMO) test (for sampling adequacy), Bartlett’s test of sphericity (for inter-item correlation significance), and the communality assessment (for the strength of factor extraction). Factor extraction was done using principal component analysis (PCA). The varimax rotation method was used for the factor solution rotation.

Factor reliability

Cronbach’s α was used as a measure of the scale’s internal consistency. Values >0.7 were considered as an indicator of good reliability.[9]

RESULTS

Sample characteristics are described in Table 1.

Table 1.

Sample characteristics

Variable	Frequency/Mean±SD
Age (in years)	44 (±14.2)
Gender
Male	51
Female	49
Marital status
Married	66
Single	22
Divorced/Separated	2
Widowed	10
Education
<5 years of education	18
5-10 years of education	28
>11 years of education	54
Religion
Hindu	79
Muslim	13
Christian	8
Occupation
Unemployed	30
Unskilled work/semiskilled work	11
Skilled/Clerical work	41
Professional work	18

Exploratory factor analysis

The scores of the three scales for the sample population (N = 100) were WHO-WBI, 53.08; ULS, 2.64; and PCS, 15.31.

The assumptions of EFA were assessed and found to be satisfactory. The adequacy of the sample confirmed using the KMO test, was 0.8, 0.74, and 0.6 for WHO-WBI, ULS, and PCS, respectively. Bartlett’s test of sphericity was found to be statistically significant (P < 0.001) for all the scales. The strength of extraction, for all the items of each scale, was >0.3. See Table 2 for factor loadings and Cronbach’s α.

Table 2.

Mean, SD, factor loadings, Cronbach’s α of items of the WHO-WBI

Item number	Mean	SD	Component factor loading	Cronbach’s α
Item 1	2.57	1.416	0.903	0.92
Item 2	2.62	1.398	0.889
Item 3	2.71	1.373	0.875
Item 4	2.81	1.468	0.857
Item 5	2.56	1.486	0.840
Mean, SD, Factor loadings, Cronbach’s α of Items of the ULS
Item 1	0.89	1.072	0.931	0.9
Item 2	0.92	1.041	0.914
Item 3	0.83	0.985	0.896
Mean, SD, Factor loadings, Cronbach’s α of Items of the PCS
Item 1	4.48	2.422	0.886	0.787
Item 2	5.91	2.775	0.859
Item 3	4.92	2.784	0.763

WHO-WBI

The results of the PCA showed that the eigenvalues were >1 for one factor in the WHO-WBI (component 1, eigenvalue: 3.810) [See Figure 1a for Scree plot]. All 5 items loaded onto a single factor, which explained 76.19% of the variance. The reliability of the extracted factor was found to be good, Cronbach’s α was 0.92.

Figure 1.

Scree plot showing the eigenvalues for the components of principal component analysis: (a) WHO-WBI, (b) ULC, (c) PCS

ULS

A similar one-factor solution was found for ULS. PCA showed that the eigenvalue was >1 for component 1, with value of 2.504 [See Figure 1b for Scree plot]. All 3 items loaded onto a single factor, which explained 83.45% of the variance. The Cronbach’s α was 0.9.

PCS

One factor in PCS (component 1, eigenvalue: 2.105) showed an eigenvalue >1 in the PCA [See Figure 1c for the scree plot]. The single factor explained 70.15% of the variance. The Cronbach’s α was 0.787.

DISCUSSION

Our study demonstrated that the Telugu versions of the WHO-WBI, ULS, and PCS exhibit fair-to-good reliability and can be explained by a single-factor solution.

WHO-WBI has been translated into 30 languages.[5] Most non-Indian language versions show a single-factor solution and high internal consistency.[10,11,12] In Indian context, the scale is translated into Bengali,[13] Gujarati,[14] and Oriya.[15] The Bengali version[15] yielded one-factor structure, which aligns with our study’s findings. The structural validity, however, was not assessed in the Gujarati[14] and Oriya[15] versions. The latent factor is, understandably, termed “well-being” in previous studies.[13] However, we deem that it may be better represented as “mood” because the two items (the ‘cheerful’ and the ‘calm’ ones) with maximum loadings for the factor assessed the mood domain. All Indian versions of the WHO-WBI so far have demonstrated good reliability, with Cronbach’s α values of 0.794 for Bengali,[13] 0.89 for Gujarati,[14] and 0.88 for Oriya.[15] Our study yielded a Cronbach’s α of 0.92 for the Telugu version, demonstrating excellent reliability.

The originally developed three-item ULS has been explained by a one-factor structure.[6] A similar result is obtained in the Japanese[16] and Greek[17] versions. In India, ULS is translated into Hindi, which has a similar one-factor structure,[18] which aligns with our results. We suppose that the latent factor that loaded onto the items accurately represents “loneliness”. The Cronbach’s α was 0.926, 0.89, and 0.75 for the Japanese, Greek, and Hindi versions, respectively.[16,17,18] Our study yielded a Cronbach’s α of 0.9, indicating similar reliability for the Telugu version.

The PCS measures critical attitudes between the patient and caregiver and plays a crucial role in predicting relapse. The Telugu scale had a reliability of 0.787, with a single factor explaining 70% of the variance. To our knowledge, this is the first validation of PCS in an Indian language. Although originally, only item 2 of the PCS was meant for assessing perceived criticism,[7] our results show that all three items are loaded by one factor that represents “criticism” correctly.

The items in the Telugu versions of three scales showed good internal consistency and can be used to measure the psychological factors related to caregiver burden in the Telugu population. Understanding of these factors can be used to predict treatment response or relapse.

Strengths and limitations

An adequate sample size for performing EFA is a key strength. The sample was collected from the two Telugu-speaking states. Scale ratings were conducted in person, which improves data accuracy and reliability. The single factor in three scales explained greater than 70% of the variance.

Lack of confirmatory factor analysis is a significant limitation. Convergent and divergent validity were not assessed. The sample was restricted to caregivers of individuals with schizophrenia, which limits the generalizability of the findings.

CONCLUSION

The Telugu versions of the three scales have fair-to-good psychometric properties.

Declaration regarding the use of generative AI

The author(s) attest that there was no use of generative artificial intelligence (AI) technology in the generation of text, figures, or other informational content of this manuscript.

Conflicts of interest

There are no conflicts of interest.

Funding Statement

Nil.