Page 79 - The value of total hip and knee arthroplasties for patients

Distinctiveness of Psychological Constructs
tests.29 The following fit indices and thresholds were used to denote a satisfactory model:Tucker-Lewis index (TLI) >0.95; comparative fit index (CFI) >0.95 and the root mean square error of approximation (RMSEA) <0.06.21 A significant Δχ2 test indicates that the model with the smallest χ2 (in this case the least stringent model) has a significantly better fit.
Model 1 hypothesised a full differentiation between the five constructs treatment
credibility, treatment expectancy, hope, optimism and pessimism. Thus items of
each construct was forced to load on a separate factor. Model 2 hypothesised
a differentiation between four constructs; the treatment credibility, treatment
expectancy and hope items were still forced to load on separate factors, but in this
model the optimism and reverse-scored pessimism items were forced to load on one 5 factor as it is controversial whether LOT-R has a uni- or bidimensional structure.18;20
Model 3 hypothesised a two factor structure in which the optimism, pessimism and hope (LOT-R and HS) items were forced to load on one factor representing ‘generalized positive beliefs about the future’ and the treatment credibility and treatment expectancy (CEQ) items were forced to load on one factor representing ‘treatment specific beliefs about the future’.This model was tested because of the theoretical plausibility that patients may have general and situational, in this case treatment specific, beliefs about the future. Model 4 hypothesised that treatment credibility, treatment expectancy, hope, optimism and pessimism items load on a single underlying latent factor.This model was tested because when it is assumed that optimism, pessimism, hope, treatment credibility and treatment expectancy are not distinguishable at all, the data should fit this one factor model. If necessary (eg because of ambiguities or high correlations between factors) post-hoc models were tested. Guttman’s lambda 2 was used to determine internal consistency reliability of each subscale.A value > 0.7 was considered indicative of good internal consistency reliability.19;35 All the analyses above were done using the total sample of THA and TKA patients.
When using the same questionnaire in different groups Factorial Invariance (FI) should be established to show that the items of the questionnaire measure the particular latent construct similarly across groups. In our study both TKA and THA patients were included, and as patients with scheduled for knee arthroplasty may face different difficulties to patients scheduled for hip arthroplasty, the constructs measured in this study may also have different meanings for these groups.
77