Skip to main content

Table 3 Analysis methods and results

From: More comprehensively measuring quality of life in life-threatening illness: the McGill Quality of Life Questionnaire – Expanded

Step Goal of analysis Procedure and criteria for model and item fit Results Conclusions
1 To ensure that new items, which were conceived to represent new domains, did not reflect existing MQOL-R domains Each of 7 candidate items (see Table 2) was added one at a time to the MQOL-R model and allowed to cross-load on all existing domains of the MQOL-R
Criteria to conclude the item belongs to an existing MQOL-R domain
• Comparative Fit Index (CFI) > 0.95
• Root Mean Square Error of Approximation (RMSEA) < 0.06
• Standardized factor loadings ≥0.6 on existing domains. Loadings between .4 and .6 were examined for conceptual fit with existing domain
Two items had standardized factor loadings > 0.4 and < 0.6:
• FINANCE on the Psychological domain (0.47)
• BURDEN on the Physical domain (0.49)
BURDEN: evidence from the qualitative interviews had suggested that this is a separate aspect of QOL from the physical domain. In addition, the loading did not reach the criterion of ≥0.6 for addition to a domain. BURDEN was therefore not added to the Physical domain
FINANCE: Conceptually different from psychological symptoms. May have loaded with them because the item captures stress due to financial situation. In addition, the loading did not reach the criterion of ≥0.6 for addition to a domain. FINANCE was therefore not added to the Psychological domain
2 Determine if each new domain individually fit with MQOL-R The 5 new domains (see Table 2) were added to the MQOL-R model one domain at a time (items were assigned to the new domains for which they were designed). For new domains that comprised a single item, the factor loading was fixed at 1.00 and the residual variance was fixed at zero.
Criteria
• CFI > 0.95
• RMSEA < 0.06
• Standardized factor loadings for all new items should be ≥ to all loadings for that item in Step 1
• CFI range: 0.96–0.97
• RMSEA range: 0.035–0.041
• Standardized factor loadings for all items for new domains were higher than in Step 1 (range: 0.72–0.88)a
All new domains fit with the existing MQOL-R, and were retained for step 3
3 Assess construct validity. Establish fit of CFA model of final version of MQOL-E Domains retained in the separate analyses in Step 2 were added all at once to MQOL-R. Another model was fit without the Finance domain
Criteria
• CFI > 0.95
• RMSEA < 0.06
Model with Finance domain
• CFI = 0.97
• RMSEA = 0.033
• Standardized factor loading range: 0.53–0.91b
Model without Finance domain
• CFI = 0.96
• RMSEA = 0.033
• Standardized factor loading range: 0.52–0.91b
The model with new domains had good fit, both with and without the Finance domain, providing support for the addition of the new domains
4 Assess whether MQOL-E domains measure a single underlying construct (overall QOL), which would support the calculation of a summary score. If so, examine the relationship of the summary score to a single item measuring global QOL
Assess importance of the Finance domain to overall QOL
The model of MQOL-E from Step 3 was fit with the addition of a second-order factor for overall QOL. This was repeated with the Finance domain removed.
The correlation of the overall factor with a single-item scale (MQOL-SIS) measure of global QOL was calculated.
Criteria
• CFI > 0.95
• RMSEA < 0.06
Model with Finance domain
• CFI = 0.93
• RMSEA = 0.042
• Standardized factor loading of Finance domain on second-order factor: 0.25
Model without Finance domain
• CFI = 0.93
• RMSEA = 0.043
• Standardized factor loadings for both first and second order factors are shown in Fig. 1
• Correlations of the factors with the MQOL-SIS are reported in Table 4
Domain means, standard deviations and internal consistency estimates (Cronbach alphas) are reported in Table 5
The second-order MQOL-E model with the Finance domain had overall acceptable model fit; although the CFI was slightly below 0.95, the RMSEA clearly meets the criterion for good fit. The loading of the Finance domain on the second-order factor was very low. The models with and without this domain had essentially the same fit (RMSEA difference of 0.001 and identical CFI). Since the Finance domain appeared to be a weak indicator of overall QOL and removing it did not result in poor model fit, the model without this domain was retained. This reduces the number of items and thus reduces participant burden.
  1. aThe Burden, Environment, and Finance factors were each represented by a single item. The loadings of these items were therefore fixed at 1.00; comparisons with Step 1 are not relevant
  2. bThis does not include the loadings for the single-item Burden, Environment and Finance factors since these loadings were fixed at 1.00