We recruited students, aged 10–12 years, from 14 elementary schools in Taiwan to test the psychometric properties of the ESQoL. Consent for the schools’ participation was obtained from the directors of student affairs through in-person visits. The students and one of their respective parents provided written informed consent. We distributed questionnaires to 953 students and 711 were returned, resulting in a response rate of 74.61%. Furthermore, for reliability testing, 60 students participated in a test–retest reliability evaluation over a 2-week period, and another 49 students participated in convergent and discriminant validity assessment.

The elementary school students’ QoL (termed ESQoL) scale was developed by employing a standardized multistep method (Barry, Chancy, Stellefson, & Chaney, 2011). We validated the scale by using a two-stage test. At Stage 1, the goal of the data reduction was to develop a comprehensive scale, which measured quality of life of students at the age of early puberty. At Stage 2, we examined the final ESQoL scale for construct validity and internal consistency.

Before the study was conducted, institutional review boards (IRBs) reviewed and approved the study protocol, informed consent, and questionnaire. We obtained informed consent from all participants included in the study. The students were provided a consent form that explained the purpose and method of the study and participants’ rights. They then shared the consent form with their parents. Participation was anonymous, confidential, and voluntary. We allowed the students and parents the freedom to decide whether to participate and provide information.

SPSS for Windows (version 18.0) was used for the statistical analyses. Data were analyzed using descriptive statistics, item analysis, EFA, and Pearson's correlation coefficient; LISREL 8.8 Software (SSI International, Inc.) was used for the CFA.

We began developing the ESQoL by applying a content analysis and constant comparison methods approach to the in-depth interview data for the focus group of fifth and sixth grade elementary students in Taiwan (Huang, Yu, Lai, & Sheu, 2007). We divided the 70 questionnaire items into four major domains (Figure 1): self-fulfillment of growth and independence (16 items); family wholeness and harmony (18 items); adaptation of identity to the school environment (20 items); and perceived social safety and stability (16 items). Items worded negatively for the construct were reverse coded and scored. The respondents rated each item on a 5-point scale, with 1=never, 2=rarely, 3=sometimes, 4=often, and 5=always.

Figure 1.

Conceptual framework of fifth and sixth grade students’ quality of life (Retrieved from Huang et al., 2007).

(0.32MB).

Content validity. The panel comprised six experts: two pediatric nursing experts, two school nurses, one qualitative research expert, and one community health expert. We provide the measurement aim, the target population, a clear framework definition, and item selection to a panel. The panel also determined the content validity index (CVI), which is used for evaluating meaning and clarity, by using a 4-point Likert scale (1=not relevant, 2=somewhat relevant, 3=quite relevant, 4=highly relevant) to avoid a neutral midpoint (Davis, 1992). The panel met thrice to evaluate the original 70 items and exclude any theoretically or practically irrelevant questions or any ambiguous items that apparently repeated the essential content of other items. We eliminated items with CVI values lower than .83, which resulted in a 50-item questionnaire. Lynn (1986) indicated that content validity depends on subjective or professional judgment; however, inclusion of at least five experts in the field determined that the I-CVI must be 1.00; when there are six experts, the I-CVI must be at least .83. Terwee et al. (2007) suggested that excellent content validity has an I-CVI of .78 or higher.

Stage 1. We examined student responses to the 50-item questionnaire in Stage 1. The goal of the data reduction was to develop a comprehensive scale, which measured quality of life of students at the age of early puberty. Further item reduction occurred in Stage 1. DeVellis (2012) suggests item reduction is a valuable means of improving the quality of items that will be used in later tests, and to assess the quality of the test as a whole. 283 students answered the 50-item preliminary questionnaire, 151 male (53.4%) and 132 female (46.4%). Participants answered the questions by responding to with one of five statements (1=never, to 5=always). The scale scores were analyzed with descriptive statistics (mean, standard deviations, slope), item difficulty measures, item discrimination indexes, and item: total correlation. Examination of the 50-item questionnaire by exploratory factor analysis (EFA) allowed for further item reduction. EFA can be employed for data reduction as it can identify items that are interrelated (DeVellis, 2012) and can eliminate redundant items, consisting of any items with similar meaning, as well as any items causing confusion. The aim of the analysis was to reveal the underlying factors and reduce the question set to retain those items that reflected a broad based definition of ESQoL.

Stage 2: Validity of the ESQoL scale. Stage 2 determined whether the final ESQoL scale accurately identified factors that are important for quality of life of elementary school students. A 24-item ESQoL questionnaire was tested in Stage 2. Validity of the questionnaire was confirmed with criterion validity, to evaluate by correlations between scores on the PedsQoL scale (Varni, Seld, & Rode, 1999) and the ESQoL, internal consistency measured and confirmatory factor analysis (CFA). In Stage 2 428 children from 5th and 6th grade completed the 24-item, 6-factor questionnaire 222 male (51.9%) and 206 female (48.1%). The CFA of the ESQoL scale assessed the relevance of the questionnaire constructed through the EFA. If the model developed from the exploratory data set has a good fit with the confirmatory data set, then the same factor structure is considered to exist in both data sets, confirming the factor structure. A maximum likelihood estimation was used to assess the model fit according to the covariance matrix of the confirmatory data set. We followed the recommendation by Browne and Cudeck (1993) and calculated the chi-square (χ2) value, the standardized root mean square residual (SRMR), and the root mean square error of approximation (RMSEA). Supplementing these indices, we examined model fit by using the comparative fit index (CFI), goodness of fit index (GFI), normed fit index (NFI), and nonnormed fit index (NNFI). Generally, the criterion for establishing model fit suggests that CFI, GFI, and NFI values close to 0.90 represent an acceptable fit and values of 0.90 or higher indicate a good fit (McCoach, Gable, & Madura, 2013).

Of the participants, 373 (52.46%) were male and 338 (47.57%) were female. 547 healthy participants, and 164 with chronic disease (Table 1). At Stage 1, using cluster random sampling, we recruited 283, 151 male (53.4%) and 132 female (46.4%), fifth and sixth grade students from six elementary schools to complete a 50-item preliminary questionnaire.

At Stage 2, we selected 428, 222 male (51.9%) and 206 female (48.1%), fifth and sixth grade students from eight elementary schools to complete a 24-item, six-factor questionnaire.

Item analysis. The means for ESQoL items ranged from 2.92 to 4.58 (SD ranged from 0.81 to 1.49). The scores for all items did not exceed 2.5 SDs from the mean, indicating that all items were adequate. In the extreme groups, the t value of each item was between 2.15 and 9.08 (p=.033 to<001); for the homogeneity test, the item-total correlation coefficients ranged from .36 to .53. On the basis of the item analysis results, we retained all items at this stage.

Exploratory factor analysis. We examined the 50-item questionnaire through the EFA, which enabled further item reduction. Following the EFA, 26 items were excluded from the 50-item scale, which resulted in 24 items and six common factors with eigenvalues greater than 1.0. The Kaiser–Meyer–Olkin measure of sampling adequacy was .81. We identified six factors, School function, Family function, Environmental life, Vitality for life, Learning ability, and Peer relationships, that explained 44.01% of the total variance (Table 2). The School function factor comprises five items for investigating the children's view of relationships with teachers and friends at school. The Family function factor comprises four items for assessing the extent of caring and interaction with family members and parents’ willingness to listen. The Environmental life factor comprises two items regarding the cleanliness, pollution, and noise level of the home environment. The Vitality for life factor comprises three items for assessing feelings about extracurricular activities and autonomy; this factor explores the children's level of leisure time and independence, which includes freedom of choice and negative reactions. The Learning ability factor comprises four items for examining feelings about academic work (difficulty in studying, comprehension of course work, and time), learning and concentration, and negative academic achievement. The Peer relationships factor comprises three items for investigating interactions with classmates, including negative relationships, acceptance by peers, and presence of friendships and bullying.

Concurrent validity. We determined the validity of the ESQoL scale by assessing criterion validity with reference to the Pediatric Quality of Life Inventory (PedsQL) generic core scale (Varni et al., 1999). Since 1998, many countries have translated, tested, and proven the validity of this scale, and it has been applied to both healthy and ill children. The self-report PedsQL gathers information from children of ages 5–18years. The four-dimensional 23-item scale comprises the domains of physical functioning (eight items), emotional functioning (five items), social functioning (five items), and school functioning (five items). Internal consistency reliability was defined as α=.09 on the original scale, or α=.62–.90 on the Chinese scale (Lin, 2012). A correlation of .30–.50 was medium; at least .70 was deemed acceptable (Cohen, 1977; Terwee et al., 2007).

In order to estimate the concurrent validity between the PedsQL and ESQoL, Pearson's correlation coefficient for the total scale (n=49) was .43 (p<.002).

Reliability. Cronbach's α was used to examine the internal consistency of the scales. A coefficient of .70 was acceptable (DeVellis, 2012). Test–retest reliability was assessed over a 2-week period in 60 students. The intraclass correlation coefficient (ICC) was used to calculate the test–retest reliability, and ICC>.70 was considered an adequate reliability score (McCoach et al., 2013). The overall scale reliability for the items was very good, with Cronbach's alpha for the subscale ranging from .74 to .90 and being .90 for the entire scale. The ICCs (test–retest reliability) ranged from .79 to .89 for the subscales and .89 for the entire scale. These results confirmed that the ESQoL has good stability.

Confirmatory factor analysis. CFA was applied to the 24 items identified in the EFA data set and provided an acceptable, but relatively poor, fit based on the goodness-of-fit statistics: χ2/df=608.17/237, p<.001, CFI=.93, GFI=.89, NFI=0.88, NNFI=.91, SRMR=.069, and RMSEA=.061. Inspection revealed that three of the 24 items had quality values lower than .50, and we excluded them. The factor structure model was retested on the remaining 21 items, and we observed support for the model in the confirmatory data set: χ2=347.27, df=174, p<001, CFI=.95, GFI=.93, NFI=.91, NNFI=.94, SRMR=.05, and RMSEA=.048 (Figure 2).

Figure 2.

Elementary school students’ quality of life model. Confirmatory factor analysis based on 21 items and six factors.

Note. School=School Function; Family=Family Function; Environment=Environmental Life; Vitality=Vitality for Life; Learning=Learning Ability; Peer=Peer Relationships; df=degrees of freedom; RMSEA=root mean square error of approximation.

(0.99MB).

We performed additional CFA to determine the goodness of fit; the ESQoL was compared with conceptualizations of five-factor structure models. Table 3 illustrates that the fit indices improved immensely when both the one-factor and uncorrelated factors models were compared with the null model. The hierarchical model was retained as the model with the best fit because it indicated that each of the six factors were interrelated first-order factors and that they were also all related to a second-order factor termed ESQoL. Retention of this model supported the assumption that these six scales are subscales of one larger scale (Figure 3).

Figure 3.

Hierarchical model and factor loadings resulting from confirmatory factor analysis.

Note. ESQoL=elementary school students’ quality of life; School=School Function; Family=Family Function; Environment=Environmental Life; Vitality=Vitality for Life; Learning=Learning Ability; Peer=Peer Relationships.

(0.76MB).

Therefore, the six scales can be administered individually or combined into one scale. Figure 3 presents the hierarchical model with factor loadings and the standardized maximum likelihood parameter estimates (path coefficients). The factor loadings and most path coefficients were strong; however, the Environmental Life factor had a weak coefficient of 0.29.

Hair, Black, Babin, Anderson, and Tatham (2006) suggested using multiple criteria when performing CFA, in addition to model fit evaluation, to gain a clear understanding of the individual parameters. We performed three additional factor extractions to confirm the model structure, as presented in Table 4: item quality (λ), composite reliability (CR), and average variance extracted (AVE). The data in Table 4 demonstrate the statistical significance that each observation variable had relative to its individual latent variable factor load (λ). All factors had values of .50 or higher, demonstrating that the observed variable sufficiently reflected its construct's latent variable (Tabachnick & Fidell, 2007). Factors with a CR of .63 were considered good (Raine-Eudy, 2000), and all factors appropriately exceeded this level, except for the Vitality for Life factor, which had a value of .63. The acceptable AVE ranged between .30 and .66; all six first-order factors demonstrated values within this range, exhibiting good strength (Fornell & Larcker, 1981).

Table 5 presents estimates for Pearson's correlation from all factors, which are similar to those the model originally proposed.

As show in Table 6, the overall of “Elementary schoolchildren quality of life” was 3.64 (SD=0.55). Analyzing from the six domains, we discovered overall quality of life have “peer relationship” rated as the highest (mean±SD=4.38±0.87), followed by “learning ability” (mean±SD=4.16±0.69), and lastly “environment life” (mean±SD=3.88±1.03). In addition, total score (t=-3.38, p<.001), school function (t=-5.70, p<.001) and vitality for life (t=-3.39, p=.001), females have significant higher than males.