Abstract
Objectives The aims of this study were to investigate the multilevel factor structure of the Therapist Presence Inventory-Therapist version (TPI-T) in a sample of Chinese beginning trainees, and to develop a 6-item brief version (TPI-T-Brief) using multilevel item response theory (M-IRT). Methods: Participants included 131 therapists from a master’s level counseling training program in China. After every session, therapists were sent the TPI-T and measures assessing their ratings of working alliance and session quality. Results: Multilevel factor analysis indicated that though conceptually TPI-T was hypothesized as having a unidimensional structure, the positively and negative worded items emerged as two statistical artifact factors. Using M-IRT, we shortened the 21-item TPI-T into the 6-item TPI-T-Brief, with three positive and three negative items. Conclusions: Evidence was found for the multilevel reliability, convergent validity, and concurrent validity of TPI-T-Brief given its significant associations with therapist- and client-rated working alliance and session quality. Further, TPI-T-Brief had significant predictive effect of session quality above and beyond working alliance at the session level. We discussed limitations (including using only one dataset) and recommended that the TPI-T-Brief be used in routine clinical practice to track therapy process.
Supplemental data
Supplemental data for this article can be accessed online at https://doi.org/10.1080/10503307.2022.2143301.
Disclosure Statement
No potential conflict of interest was reported by the author(s).
Notes
1 This notation of “2w-3b” means a model with 2 factors at the within level and 3 factors at the between level. Same for 3w-3b below.
2 To note, for the M-IRT analyses, we used the original, rather than reverse-coded, response points for the negatively worded items, with higher points indicating lower therapist presence. The positively worded items also used their original response points with higher points indicating higher therapist presence.
3 According to scholars (e.g., Pallant, Citation2011), it is common to obtain small reliability values (e.g., around .50) when the number of items for a scale is small (e.g., less than 10 items). Thus, we used Spearman-Brown correction to adjust for the impact of the small number of items.