Psychometric properties of the Tinnitus Functional Index (TFI): Assessment in a UK research volunteer population

OBJECTIVES Questionnaires are essential for measuring tinnitus severity and intervention-related change but there is no standard instrument used routinely in research settings. Most tinnitus questionnaires are optimised for measuring severity but not change. However, the Tinnitus Functional Index (TFI) claims to be optimised for both. It has not however been fully validated for research purposes. Here we evaluate the relevant psychometric properties of the TFI, specifically the questionnaire factor structure, reproducibility, validity and responsiveness guided by quality criteria for the measurement properties of health-related questionnaires. METHODS The study involved a retrospective analysis of data collected for 294 members of the general public who participated in a randomised controlled trial of a novel tinnitus device (ClinicalTrials.gov Identifier: NCT01541969). Participants completed up to eight commonly used assessment questionnaires including the TFI, Tinnitus Handicap Inventory (THI), Tinnitus Handicap Questionnaire (THQ), a Visual Analogue Scale of loudness (VAS-Loudness), Percentage Annoyance question, the Beck's Depression Inventory (BDI), Beck's Anxiety Inventory (BAI), and the World Health Organisation Quality of Life-Bref (WHOQOL-BREF). A series of analyses assessed the study objectives. Forty four participants completed the TFI at a second visit (within 7-21 days and before receiving any intervention) providing data for reproducibility assessments. RESULTS The 8-factor structure was not fully confirmed for this general (non-clinical) population. Whilst it was acceptable standalone subscale, the 'auditory' factor showed poor loading with the higher order factor 'functional impact of tinnitus'. Reproducibility assessments for the overall TFI indicate high internal consistency (α = 0.80) and extremely high reliability (ICC: 0.91), whilst agreement was borderline acceptable (93%). Construct validity was demonstrated by high correlations between scores on the TFI and THI (r = 0.82) and THQ (r = 0.82), moderate correlations with VAS-L (r = 0.46), PR-A (r = 0.58), BDI (r = 0.57), BAI (r = 0.39) and WHOQOL (r = -0.48). Floor effects were observed for more than 50% of the items. A smallest detectable change score of 22.4 is proposed for the TFI global score. CONCLUSION Even though the proposed 8-factor structure was not fully confirmed for this population, the TFI appears to cover multiple symptom domains, and to measure the construct of tinnitus with an excellent reliability in distinguishing between patients. While the TFI may discriminate those whose tinnitus is not a problem, floor effects in many items means it is less appropriate as a measure of change in this subgroup. Further investigation is needed to determine whether these effects are relevant in other populations.