One for One or One for All? A Comparison of the Tinnitus Functional Index and the Client Oriented Scale of Improvement in Tinnitus as Outcome Measures

Abstract

Background:

The purpose of the Client Oriented Scale of Improvement in Tinnitus (COSIT) is to set treatment goals and assess individual outcomes. The Tinnitus Functional Index (TFI) was developed as a comprehensive questionnaire to measure individual and population responses to tinnitus therapies.

Purpose:

To investigate the convergent validity and responsiveness of the COSIT and TFI.

Study Sample:

A secondary data analysis of COSIT and TFI measures from 99 participants who used sound-based therapy with hearing aids or earphones was undertaken.

Data Collection and Analysis:

Athematic analysis of COSIT goals was applied. Nonparametric statistical methods were used to measure convergence of COSIT “change with treatment scores” and “final scores” to change in TFI scores pretreatment and posttreatment and the final TFI scores with treatment. The level of agreement between the measures was determined using receiver operating characteristic curves. Treatment responsiveness was determined by comparing treatment-related change in total scores and subscales for responders versus nonresponders. Responders were classified as having >13-point change in the total TFI score and COSIT “change” score of >2 (3 = slightly better to 5 = much better).

Results:

The five most common COSIT goals were reducing tinnitus impact by addressing (1) well-being, (2) hearing, (3) sleep, (4) focus, and (5) context. The TFI and COSIT change scores and TFI and COSIT final scores showed moderate to strong agreement. The COSIT classified a higher percentage (78 percent) of participants as responders than the TFI did (43 percent). There was a fair agreement between the clinically meaningful change scores for TFI and COSIT (κ = 0.29).

Conclusions:

The TFI and COSIT measured similar constructs, and both differentiated responders from nonresponders. Our results are consistent with the COSIT being an ideal “one for one” tool because of its high responsiveness to individual goals. The TFI is better suited as a “one for all” questionnaire because of its universal content. Clinical trials should consider use of both the COSIT and TFI because of their different, but complementary, strengths.

INTRODUCTION

Tinnitus, subjective ear and head sound, is heterogenous in presentation and effect (Cederroth et al., 2024), and there is no gold standard objective measure or biomarker for its presence (Kang et al., 2021). Psychometric matching can characterize the nature of the tinnitus sound, but measures such as loudness are weakly correlated to how much impact tinnitus has on a person (Hall et al., 2016; Vajsakovic et al., 2021). For these reasons, clinical assessment and research have largely relied on questionnaires to measure the outcomes of treatment (Theodoroff, 2021).

Tinnitus questionnaires are used to determine the severity or handicap associated with tinnitus. Questionnaires such as the Tinnitus Functional Index (TFI) (Meikle et al., 2012), Tinnitus Questionnaire (Goebel & Hiller, 1994), Tinnitus Handicap Inventory (THI) (Newman et al., 1996), and Tinnitus Primary Function Questionnaire (Tyler et al., 2014) are meaningful for most people and can be generalized to different populations. The TFI consists of 25 questions and eight subscales measuring different tinnitus effects: (1) intrusiveness, (2) sense of control, (3) cognitive, (4) sleep, (5) auditory, (6) relaxation, (7) quality of life, and (8) emotion. The TFI evaluates different aspects of tinnitus handicap more evenly than several of the other questionnaires (Fackrell & Hoare, 2014). The TFI has been used to show changes in tinnitus distress over time and has been shown to be responsive to treatment in several languages and countries (Boecking et al., 2021; Chandra et al., 2018; Jacquemin et al., 2019). The change in TFI scores that represents a clinically meaningful change with treatment has been estimated to be as low as 4.8 (Chandra et al., 2018) and as high as 22 (Fackrell et al., 2016). A reduction in TFI of 13 points was suggested as a minimal important change by the originators of the TFI (Meikle et al., 2012). Subsequently, a change of greater than 13 points has become the de facto TFI responder criteria, despite debate whether this is appropriate (Fackrell et al., 2022).

The TFI has proven validity and measures multiple domains but like most tinnitus questionnaires, it uses a predetermined closed set of questions. It is possible that some or many TFI questions may not be relevant to some people. An alternative approach is to use an open-set approach in which the client chooses their own issues to which therapy success is assessed. The Client Oriented Scale of Improvement in Tinnitus (COSIT) is an open-set questionnaire primarily used for goal planning and assessment of outcomes posttherapy (Searchfield, 2019). The COSIT shares similarities with Goal-Attainment Scaling (Wagenaar et al., 2024) and is based on a commonly used hearing aid outcome measure (the Client Oriented Scale of Improvement) (Dillon et al., 1997). The clinician can use the scale to consider the client’s needs and develop a therapy plan accordingly. Outcomes are measured against the five most relevant ranked problems for the individual. Searchfield (2019) evaluated the COSIT in a sample of 122 participants and identified 11 themes, with the six most common needs for therapy being (1) hearing, (2) well-being, (3) ability to cope, (4) understanding tinnitus, (5) sleep, and (6) effect of context on tinnitus. The COSIT had moderate convergent validity with the THI, Tinnitus Handicap Questionnaire, and TFI, and it was concluded that it was a valid method for measuring tinnitus therapy outcomes. The value of the COSIT for clinical use has been acknowledged, but its applicability as a clinical trial outcome has been questioned (Thompson et al., 2024). A tailored outcome measure for each participant in a study would limit comparison of group effects. Unlike the TFI, participants completing the COSIT would report intervention outcomes to their unique problems; the outcomes measured would not be uniform between participants or groups. The absence of standardization would make group comparisons difficult (Thompson et al., 2024). The COSIT’s responsiveness and distribution of responders and nonresponders have also not yet been compared with the other questionnaires.

In this study, we compare the COSIT with the TFI to determine whether they measure the same constructs and differentiate between persons who benefit or do not benefit from therapy. We replicate studies undertaken to compare the TFI’s convergent validity and responsiveness with other questionnaires (Boecking et al., 2021; Jacquemin et al., 2019). We determine the convergent validity of the change in TFI and posttreatment TFI scores with COSIT change in tinnitus and final scores. TFI and COSIT minimal clinically important difference criteria were used to ascertain responsiveness to therapy. Change in responder versus nonresponder scores with therapy and agreement between the questionnaires’ responder classifications were determined. Boecking et al.’s (2021) equivalency study of common closed-set tinnitus questionnaires had the creative title of “All for One and One for All?” Our study reimagines that title to describe our investigation, using similar methods, but applied to the outcomes of the open-set COSIT with outcomes to individual goals (“one for one”) compared with the TFI as a universal standard closed-set questionnaire for measuring therapy outcome (“one for all”). Implications of the study are recommendations as to if and how the questionnaires should be used clinically and in research.

METHODS

This is a secondary analysis of data extracted from two trials (Australian New Zealand Clinical Trials Registry; ACTRN12621000389808 approved by the University of Auckland Human Participants Ethics Committee and trial ACTRN12621001754831 approved by the Northern B Health and Disability Ethics Committee New Zealand). All participants provided written informed consent in accordance with the Declaration of Helsinki.

Participants

The analysis sample consisted of n = 99 adult research participants who attended the University of Auckland tinnitus research laboratory between 2020 and 2023. Participants completed COSIT goals and answered TFI questions at baseline, and then scored COSIT goals and answered TFI questions 3 months after sound therapy of different forms using hearing aids or earphones (Sanders et al., 2023; Searchfield & Sanders, 2022). The 57 male and 42 female participants were aged between 23 and 91 years old. The participants all reported experiencing chronic tinnitus for more than 3 months. Participant characteristics are documented in the Appendix.

TFI

The TFI consisted of 25 items and eight subscales scored on a 0- to 10-point Likert scale. Participants were asked to report their impressions of tinnitus based on the past week. The TFI scores were converted to a score out of 100. The total TFI score was the sum of the responses obtained from all questions, divided by the number of questions answered, multiplied by 10. The eight subscales relating to tinnitus function effects were (1) intrusiveness, (2) sense of control, (3) cognitive, (4) sleep, (5) auditory, (6) relaxation, (7) quality of life, and (8) emotion. Median scores were calculated for the total TFI and subscores. A change score of at least 13 points was used as a threshold for clinically meaningful improvement (Meikle et al., 2012).

COSIT

Participants were asked to identify up to five goals for tinnitus treatment. Once treatment goals were identified, participants were asked to rank the priority of their goals for therapy (from most important [1] to least important [5]). After therapy the participants were asked about any change in their tinnitus: “With the therapy your tinnitus is (1) worse, (2) no different, (3) slightly better, (4) better, or (5) much better?” For the final score with tinnitus, participants were asked: “With tinnitus therapy you are annoyed by your tinnitus (1) almost always, (2) most of the time, (3) half of the time, (4) occasionally, or (5) hardly ever?” Responses were marked in the correct column on the COSIT questionnaire form. The change scores classified as responders were endorsed by participants as follows: 3 = slightly better, 4 = better, or 5 = much better. The final scores were classified as often problem (1 = almost always, 2 = most of the time) or less often problem (3 = half of the time, 4 = occasionally, and 5 = hardly ever). The median degree of change (“change score,” across goals) and the median final result with therapy (“final score,” across goals) were calculated. The ranking of goals was not considered in the analysis.

Analysis

To enable comparison with the TFI at a population level, we organized individual COSIT goals into themes using thematic analysis (stage A, step 2; Attride-Stirling, 2001). The initial thematic analysis was undertaken by one of the authors (V.K.). Individual goals were recorded and sorted according to commonality (e.g., “I want to sleep uninterrupted,” “Reduce tinnitus when trying to get to sleep”). Goals were allocated to one theme only; ambiguous or items that could be allocated to multiple themes were marked for review. The allocation to themes was reviewed by one of the authors (G.D.S.), and a consensus on categories and allocation was agreed on by the authors. After cataloguing responses, those themes endorsed by four or fewer participants were not included in further analysis because the number of responses was considered too few for generalization and lacked sufficient sample size for inclusion in statistical analysis. Analysis of data and graphing was undertaken using IBM SPSS 29.0.2.0 (Sanders et al., 2023) and GraphPad Prism 9.3.1 for Apple MacIntosh unless otherwise stated. The frequency distributions of the TFI and COSIT were examined and found to be nonnormally distributed, so nonparametric statistics were used. TFI scores before and after therapy were compared using the Mann-Whitney test. The Cohen effect size for the Mann-Whitney U statistic was calculated (Lenhard & Lenhard, 2022).

Convergent Validity

Convergent validity between TFI and COSIT was undertaken using the Kendall tau statistic. The Kendall tau, a measure of rank correlation, is a nonparametric substitution for intraclass correlation (weak positive agreement, tau > 0.09; moderate, tau > 0.27; strong, tau > 0.49) (Wicklin, 2023).

Responsiveness

Participants were classified, independently, for each questionnaire as responders if they had a >13-point change in the total TFI score with therapy and >2-point change in the COSIT change score (3 = slightly better, 4 = better, 5 = much better). Participants with TFI changes ≤13 and ≤2 for the COSIT change score were classified as nonresponders. Kappa (κ) coefficients were calculated to compare the agreement in responder classifications (quickcalcs, www.graphpad.com).

The total and five most common goal COSIT change scores were compared with the change in total and subscale TFI scores (difference before and after therapy) using the Kruskal-Wallis test with Dunn’s correction for multiple comparisons. The COSIT “final” results (time annoyed by tinnitus for the five most common goals) were similarly compared with the final TFI scores and subscales after therapy. Receiver operator characteristics (ROC) analyses were undertaken to ascertain whether change or final scores effectively distinguished between responders and nonresponders for each questionnaire. The criteria for responders and nonresponders from one questionnaire was used to group data from the other questionnaire. Then scores classified as responders are compared with nonresponder scores. COSIT scores were grouped according to the TFI criteria and COSIT criteria and vice versa. The ROC curve plots probability of a score correctly differentiating responder and nonresponders. The “area under the curve” (AUC) statistic was calculated to enable comparison between the TFI and COSIT classification of responders.

RESULTS

Participant Characteristics

The participant characteristics are summarized in the Appendix. Participants were on average 57 years old (standard deviation [SD] 13.8). The average TFI was 52.3 (SD 18.1), a moderate to severe tinnitus impact. Baseline TFI scores were skewed toward higher tinnitus impact (Figure 1). Self-reported tinnitus loudness was 69.4 (SD 15.2) out of a maximum of 100. Participants commonly had hearing loss (83 percent), but few reported psychiatric disorders (7 percent). TFI total scores showed significant improvement between baseline and posttreatment median scores (16 points, Mann-Whitney U = 3152, p < 0.0001, Cohen d = 0.6).

Figure 1.

Frequency distributions of the total scores of the TFI measure at baseline.

Treatment Goal Categorization

The five most common goal themes (Table 1) were (1) well-being (improve how I feel and be less depressed or anxious); (2) hearing (reduce the interference of tinnitus on communication); (3) sleep (reduce tinnitus impact on sleep and improve sleep quality); (4) focus (reduce attention to tinnitus and its negative effect on cognition); and (5) context (manage tinnitus, i.e., reactive to the environmental, e.g., worse in quiet).

Table 1.

Number of Participants, Mean (SD), and Median Scores for the COSIT and TFI

COSIT
	Change scores			Final scores
	n	Mean (SD)	Median	n	Mean (SD)	Median
Total	99	2.86 (0.87)	2.67	99	3.34 (1.00)	3.50
Well-being	72	2.94 (1.03)	3.00	72	3.30 (1.11)	3.58
Hearing	66	2.91 (0.99)	3.00	66	3.55 (1.11)	4.00
Sleep	60	2.81 (0.98)	2.25	60	3.11 (1.14)	3.00
Focus	52	2.72 (0.99)	2.00	52	3.31 (1.22)	3.00
Context	48	2.79 (0.95)	3.00	49	3.17 (1.16)	3.50
Cope	4	3.25 (0.96)	3.50	4	3.00 (0.00)	3.00
TFI
	Change in scores			Final scores
	Number	Mean (SD)	Median	n	Mean (SD)	Median
Total	96	11.63 (20.53)	11.40	96	40.40 (21.53)	38.00
Intrusive	96	9.58 (22.56)	5.00	96	49.83 (21.9)	50.00
Sense of control	96	11.81 (25.76)	10.00	96	49.41 (24.86)	50.00
Cognitive	96	10.31 (22.84)	10.00	96	33.19 (23.61)	26.67
Sleep	96	12.01 (22.96)	10.00	96	33.30 (27.91)	30.00
Auditory	96	14.51 (25.63)	13.33	96	44.44 (25.83)	45.00
Relaxation	96	10.66 (27.71)	8.33	96	49.41 (27.72)	50.00
Quality of life	96	11.46 (22.98)	10.00	96	32.66 (24.38)	27.50
Emotional	96	12.78 (24.92)	10.00	96	33.54 (26.65)	26.67

Note: The COSIT is completed once after therapy. The change in TFI is the difference between pretherapy and posttherapy questionnaires. The TFI final score is the TFI posttherapy.

Convergent Validity

The Kendall tau, a measure of rank correlation, measured consistency in total and subscale scores between (1) the TFI change (before and after therapy) with the COSIT change score (Table 2) and (2) the final TFI score and COSIT final problem score (Table 3). The total change scores showed a moderate-strong tau (0.33). For the final total scores, the tau score was higher (0.39), indicating a stronger rank correlation. All subscales for the change measurements were significantly correlated, except TFI intrusive with COSIT well-being and TFI quality of life with COSIT focus. The strongest agreement in score ranks was between change in the TFI auditory scale and COSIT sleep scale (tau = 0.47). For final scores, all tau values were significant except TFI auditory with COSIT focus and TFI quality of life with COSIT sleep.

Table 2.

Convergent Validity of COSIT Change Scores and Change in TFI with Therapy

	TFI
COSIT	Total	Intrusive	Sense of control	Cognitive	Sleep	Auditory	Relaxation	Quality of life	Emotional
Total	0.33***	0.23**	0.28***	0.29***	0.27***	0.33***	0.31***	0.23**	0.31***
	(0.21–0.45)	(0.10–0.35)	(0.16–0.40)	(0.16–0.41)	(0.14–0.39)	(0.21–0.45)	(0.19–0.43)	(0.09–0.35)	(0.19–0.43)
Well-being	0.37***	0.28**	0.28**	0.34***	0.30**	0.40***	0.33***	0.27**	0.31***
	(0.23–0.50)	(0.13–0.42)	(0.12–0.42)	(0.20–0.48)	(0.15–0.44)	(0.26–0.52)	(0.18–0.46)	(0.10–0.41)	(0.16–0.45)
Hearing	0.29**	0.110	0.27**	0.20*	0.24*	0.32***	0.28**	0.25**	0.32***
	(0.14–0.44)	(−0.05 to 0.27)	(0.11–0.41)	(0.04–0.35)	(0.08–0.39)	(0.170–0.46)	(0.13–0.43)	(0.09–0.40)	(0.17–0.46)
Sleep	0.39***	0.34**	0.27*	0.40***	0.35***	0.47***	0.36**	0.31**	0.35**
	(0.23–0.54)	(0.18–0.49)	(0.10–0.42)	(0.24–0.54)	(0.18–0.49)	(0.32–0.59)	(0.12–0.51)	(0.14–0.46)	(0.19–0.500)
Focus	0.33**	0.37***	0.30**	0.37***	0.27*	0.29*	0.34**	0.120	0.31**
	(0.16–0.49)	(0.20–0.52)	(0.12–0.46)	(0.20–0.52)	(0.09–0.44)	(0.10–0.45)	(0.16–0.50)	(−0.07 to 0.30)	(0.13–0.47)
Context	0.38***	0.25*	0.32**	0.29**	0.35**	0.26*	0.36**	0.24*	0.39**
	(0.21–0.54)	(0.06–0.42)	(0.14–0.48)	(0.11–0.46)	(0.17–0.51)	(0.07–0.43)	(0.19–0.52)	(0.05–0.41)	(0.21–0.54)

Note: Scores are Kendall’s tau (95% CI). Strong positive agreement between the ranks when tau > 0.45, moderate positive agreement between the ranks when tau > 0.27, and weak positive agreement between the ranks when tau > 0.09.

Significance: *p < 0.05, **p < 0.01, ***p < 0.001.

Table 3.

Convergent Validity of COSIT Final Scores and TFI Posttherapy

	TFI
COSIT	Total	Intrusive	Sense of control	Cognitive	Sleep	Auditory	Relaxation	Quality of life	Emotional
Total	−0.39***	−0.36***	−0.41***	−0.40***	−0.32***	−0.24***	−0.38***	−0.31***	−0.39***
	(−0.5 to −0.27)	(−0.47 to −0.24)	(−0.520 to −0.300)	(−0.50 to −0.28)	(−0.43 to −0.19)	(−0.37 to −0.11)	(−0.49 to −0.26)	(−0.43 to −0.18)	(−0.49 to −0.27)
Well-being	−0.39***	−0.41***	−0.42***	−0.36***	−0.25**	−0.35***	−0.39***	−0.33***	−0.34***
	(−0.52 to −0.25)	(−0.53 to −0.27)	(−0.54 to −0.27)	(−0.49 to −0.22)	(−0.40 to −0.10)	(−0.48 to −0.20)	(−0.52 to −0.24)	(−0.46 to −0.18)	(−0.47 to −0.19)
Hearing	−0.31**	−0.28**	−0.26**	−0.36***	−0.33***	−0.27**	−0.25**	−0.28**	−0.31**
	(−0.45 to −0.15)	(−0.42 to −0.12)	(−0.41 to −0.10)	(−0.50 to −0.21)	(−0.47 to −0.17)	(−0.41 to −0.12)	(−0.40 to −0.09)	(−0.43 to −0.12)	(−0.46 to −0.16)
Sleep	−0.29**	−0.29**	−0.37***	−0.34***	−0.23*	−0.20*	−0.26*	−0.18	−0.31**
	(−0.45 to −0.12)	(−0.45 to −0.12)	(−0.51 to −0.20)	(−0.49 to −0.18)	(−0.39 to −0.06)	(−0.37 to −0.03)	(−0.41 to −0.09)	(−0.35 to −0.01)	(−0.46 to −0.15)
Focus	−0.36***	−0.31**	−0.37***	−0.37**	−0.33**	−0.19	−0.41***	−0.26*	−0.37***
	(−0.51 to −0.18)	(−0.47 to −0.13)	(−0.53 to −0.20)	(−0.52 to −0.19)	(−0.49 to −0.15)	(−0.36 to 0.00)	(−0.56 to −0.24)	(−0.43 to −0.08)	(−0.53 to −0.20)
Context	−0.39***	−0.35**	−0.50***	−0.35**	−0.27*	−0.16	−0.37**	−0.37**	−0.33**
	(−0.54 to −0.21)	(−0.51 to −0.17)	(−0.63 to −0.34)	(−0.51 to −0.17)	(−0.44 to −0.09)	(−0.34 to 0.03)	(−0.52 to −0.19)	(−0.52 to −0.19)	(−0.49 to −0.14)

Note: Scores are Kendall’s tau (95% CI). Strong positive agreement between the ranks when tau > 0.45, moderate positive agreement between the ranks when tau > 0.27, and weak positive agreement between the ranks when tau > 0.09.

Significance: *p < 0.05, **p < 0.01, ***p < 0.001.

Responders Versus Nonresponders Classification

Following the criteria developed for each questionnaire, the COSIT identified 78 percent of participants as responders and the TFI identified 43 percent as responders. Responsiveness was determined for each scale using minimum meaningful scores, >13-point change in the TFI scale before and after therapy and >2 on the COSIT change score (Table 4). These criteria were then applied to the overall scores, TFI subscales, and COSIT goals for change scores and final scores. The criteria for meaningful change separated responders and nonresponders for the change (Figure 2) and final scores (Figure 3) to different degrees.

Table 4.

Median Scores for Responders Versus Nonresponders as Classified by Each Tinnitus Questionnaire

COSIT change score							COSIT final score
	COSIT >2			TFI >13			COSIT >2			TFI >13
	Responder	Nonresponder	p	Responder	Nonresponder	p	Less often	More often	p	Responder	Nonresponder	p
Total	3.00	2.00	<0.001	3.00	2.00	>0.999	3.75	2.00	<0.001	4.00	3.00	0.02
Hearing	3.00	2.00	<0.001	3.00	2.00	>0.999	4.00	2.00	<0.001	4.00	4.00	>0.999
Well-being	3.00	2.00	<0.001	3.00	2.00	>0.999	4.00	2.00	<0.001	4.00	3.00	0.01
Context	3.00	2.00	<0.001	3.00	2.00	>0.999	4.00	2.00	<0.001	4.00	2.50	0.11
Sleep	3.00	2.00	<0.001	3.00	2.00	>0.999	3.00	2.00	<0.001	3.50	3.00	0.21
Focus	3.00	2.00	<0.001	3.00	2.00	>0.999	4.00	2.00	<0.001	4.00	3.00	>0.01

Change in TFI score							TFI final score
	COSIT >2			TFI >13			COSIT >2			TFI >13
	Responder	Nonresponder	p	Responder	Nonresponder	p	Less often	More often	p	Responder	Nonresponder	p
Total	15.40	1.00	0.04	24.80	−1.80	<0.001	37	61.8	0.07	25.60	51.20	<0.001
Intrusive	6.67	0.00	0.10	26.67	−3.33	<0.001	50	68.33	0.25	36.67	60.00	<0.001
Sense of control	13.33	−1.67	<0.001	23.33	0.00	<0.001	46.67	81.67	0.01	36.67	63.33	<0.001
Cognitive	11.67	5.00	0.38	26.67	0.00	<0.001	26.67	61.67	0.03	20.00	40.00	<0.001
Sleep	11.67	6.67	0.19	26.67	3.33	<0.001	30	56.67	0.24	16.67	40.00	<0.001
Auditory	20.00	0.00	0.02	30.00	0.00	<0.001	40	58.33	>0.99	30.00	60.00	<0.001
Relaxation	10.00	−3.33	0.03	33.33	−6.67	<0.001	45	81.67	0.05	30.00	70.00	<0.001
Quality of life	12.50	1.25	0.58	30.00	−1.25	<0.001	22.5	58.75	0.06	20.00	42.50	<0.001
Emotional	13.33	5.00	0.36	26.67	−1.67	<0.001	23.33	66.67	0.02	16.67	40.00	<0.001

Note: Gray shading highlights statistically significant differences.

Figure 2.

Frequency distributions of the total scores, for TFI difference before and after therapy, for COSIT “change” score. (A) COSIT with COSIT >2 responder criteria. (B) COSIT with TFI >13 criteria. (C) TFI with TFI >13 criteria. (D) TFI with COSIT >2 criteria.

Figure 3.

Frequency distributions of the total TFI final scores after therapy and COSIT “final result.” (A) COSIT with COSIT >2 responder criteria. (B) COSIT with TFI >13 criteria. (C) TFI with TFI >13 criteria. (D) TFI with COSIT >2 criteria.

Change with Therapy

A Kruskal-Wallis test revealed a significant difference in COSIT change scores when the >2 COSIT change score criteria (determined after therapy) were applied to the COSIT to classify responders (n = 72) and nonresponders (n = 27) [H(6) = 166.6, p < 0.0001]. Post hoc Dunn’s tests with Bonferroni correction indicated the following significant pairwise comparisons between responders and nonresponders: overall COSIT change (p < 0.0001; Figure 2A), hearing (p < 0.0001), well-being (p < 0.0001), context (p < 0.001), sleep (p < 0.0001), and focus (p < 0.0001) goals. A Kruskal-Wallis test revealed a significant difference in COSIT scores when the >13-point change in TFI criteria (before vs. after questionnaires) was applied to the COSIT to classify responders (n = 43) and nonresponders (n = 53) [H(6) = 141.4, p < 0.0001]. Post hoc Dunn’s tests with Bonferroni correction indicated no difference in median scores for the overall change score (Figure 2B) and goals. A Kruskal-Wallis test revealed a significant difference in TFI scores when the >13-point TFI change score criteria were applied to classify responders (n = 43) and nonresponders (n = 56) to change in TFI scores [H(8) = 403.5, p < 0.0001]. Post hoc Dunn’s tests with Bonferroni correction indicated the following significant pairwise comparisons between responders and nonresponders: overall change in TFI (p < 0.0001; Figure 2C), intrusive, sense of control, cognitive, sleep, auditory, relaxation, quality-of-life, and emotion subscales (p < 0.0001). A Kruskal-Wallis test revealed a significant difference in TFI scores when the >2 COSIT change score criteria were applied to classify responders (n = 70) and nonresponders (n = 26) [H(8) = 89.63, p < 0.0001]. Post hoc Dunn’s tests with Bonferroni correction indicated the following significant pairwise comparisons between responders and nonresponders: overall change in TFI (p < 0.05; Figure 2D), sense of control (p < 0.001), auditory (p < 0.05), and relaxation (p < 0.05). Intrusive, cognitive, sleep, quality-of-life, and emotion subscale median scores for change in TFI were not significantly different between responders and nonresponders when using the COSIT >2 criteria. A Kruskal-Wallis test revealed a significant difference in TFI scores when the >13 TFI change score criteria were applied to classify responders (n = 46) and nonresponders (n = 53) with regard to their final TFI totals [H(8) = 259.4, p < 0.0001]. Post hoc Dunn’s tests with Bonferroni correction indicated the following significant pairwise comparisons between responders and nonresponders: overall change in TFI (p < 0.0001; Figure 2C), intrusive, sense of control, cognitive, sleep, auditory, relaxation, quality-of-life, and emotion subscales (p < 0.0001).

Final Scores

The final scores consist of the TFI after therapy and the COSIT rating of final degree of problem. A Kruskal-Wallis test revealed a significant difference in COSIT final scores when the >2 COSIT final score criteria (n = 14 often problem, n = 85 less often problem) were applied to the COSIT [H(6) = 171.2, p < 0.0001]. Post hoc Dunn’s tests with Bonferroni correction indicated the following significant pairwise comparisons between responders and nonresponders: overall COSIT change (p < 0.0001; Figure 3A), hearing (p < 0.0001), well-being (p < 0.001), context (p < 0.001), sleep (p < 0.001), and focus (p < 0.0001) goals. A Kruskal-Wallis test revealed a significant difference in COSIT final scores when the >13-point TFI criteria were applied (n = 46 responders, n = 53 nonresponders) to the COSIT [H(6) = 51.94, p < 0.0001]. Post hoc Dunn’s tests with Bonferroni correction indicated the following significant pairwise comparisons between responders and nonresponders. The overall COSIT change (p < 0.05; Figure 3C), well-being (p < 0.05), and focus (p < 0.005) goals were significantly different between responders and nonresponders. The hearing, context, and sleep goal median scores were not significantly different when the TFI >13 responder criteria were applied to final COSIT scores. A Kruskal-Wallis test revealed a significant difference in TFI scores when the >13 TFI change score criteria were applied to classify responders (n = 43) and nonresponders (n = 53) to final TFI scores [H(8) = 259.4, p < 0.0001]. Post hoc Dunn’s tests with Bonferroni correction indicated the following significant pairwise comparisons between responders and nonresponders: overall change in TFI (p < 0.0001; Figure 3C) sense of control, relaxation, and emotion subscales (p < 0.0001) and cognitive, sleep, quality-of-life (p < 0.0005), intrusive, and auditory (p < 0.005) subscales. A Kruskal-Wallis test revealed a significant difference in TFI final scores when the >2 COSIT final score criteria (n = 14 often problem, n = 82 less often problem) were applied to the TFI [H(8) = 171.2, p < 0.0001]. Post hoc Dunn’s tests with Bonferroni correction indicated the following significant pairwise comparisons between responders and nonresponders: sense of control (p < 0.01), cognitive, relaxation, and emotion subscales (p < 0.05). The final TFI score (Figure 3D) intrusive, sleep, auditory, and quality-of-life subscales did not differ when classified using the >2 COSIT final score criteria.

Agreement Between the Measures’ Responders Versus Nonresponders Classifications

Comparisons of the questionnaires’ responders classifications revealed fair agreement (64.65 percent of the observations) between the COSIT and TFI overall scores (κ = 0.29, 95 percent confidence interval [CI]: 0.161–0.424).

ROC Analyses

ROC AUC analyses were used to ascertain the ability of the COSIT and TFI to distinguish between responders and nonresponders in the other questionnaire. The change in TFI score weakly distinguished between responders and nonresponders in the COSIT change score (AUC, 0.71, 95 percent CI: 0.60–0.81) and COSIT final score (0.67, 0.57–0.78). The COSIT score criteria of >2 points weakly distinguished between responders and nonresponders with change in TFI score (0.69, 0.58–0.80) but did not distinguish for the final TFI score (0.52, 0.39–0.66).

DISCUSSION

The convergent validity and responsiveness of the COSIT and the TFI were compared after sound-based tinnitus therapy. The questionnaires were completed by 99 participants with moderate-severe chronic tinnitus. The intervention of sound therapy using hearing aids or earphones resulted in a TFI change with a moderate-large effect size. The TFI and COSIT final scores were moderately correlated, showing slightly higher correlation than change in TFI and COSIT change scores. Most of the COSIT goals were significantly correlated with TFI subscales. The criteria of >13-point change in TFI and >2 in COSIT change score differentiated responders and nonresponders when applied to the origin questionnaires. When the responsiveness criteria were crossed over to the comparison questionnaire (applying TFI criteria to COSIT and vice versa), there was no significant difference in COSIT change scores, indicating the TFI criterion did not differentiate COSIT responders or nonresponders. The COSIT criteria did differentiate total TFI and sense of control, auditory, and relaxation subscale responders and nonresponders. The Kappa statistic indicated a fair agreement in responsiveness between the TFI and COSIT. A ROC analysis of AUC found the TFI score of >13 weakly distinguished between responders and nonresponders in the COSIT change score, and the COSIT score criteria of >2 weakly distinguished responders in change in TFI score.

The final TFI scores represent the status of tinnitus after the intervention, accordingly they were lower than baseline, but the final score depends on both baseline and degree of change. The COSIT final score was an indicator of how much a problem tinnitus was and could feasibly be the same as before therapy. The TFI responder criteria differentiated COSIT final total, well-being, and focus scores. The TFI score weakly distinguished between responders and nonresponders in the final COSIT score. Using the COSIT criteria, there was no difference in total final TFI scores. TFI scores after treatment were evenly distributed according to COSIT responder status (responder vs. nonresponder). The subscales of sense of control, cognitive, relaxation, and emotion did significantly differ. At a population level the TFI score after therapy and the final COSIT score with therapy do not indicate treatment benefit per se. The impact of treatment could be greater in an individual moving from a TFI score of 50 to 25 than from 25 to 20. The scores do provide insight whether further improvement is possible. A halving of tinnitus would be viewed as a good therapeutic outcome by researchers and clinicians; however, we should not lose sight that the tinnitus that remains can still have a large impact on the person’s life. Clinicians should address these final scores with clients and plan the next therapy steps to try to further reduce residual tinnitus. In the case of the COSIT, the goals and their priority can be reassessed. Goals may fall out of the top five COSIT needs and be replaced by new goals. This adaptive process enables continued improvement and encourages the client-clinician therapeutic alliance as a collaborative process. There is a need for an iterative therapeutic process in the absence of “single-shot” tinnitus cures.

The TFI subscales and most common COSIT goals overlapped but were not always strongly correlated. For example, change in the TFI sleep subscale was most strongly correlated with change score to the COSIT goal of improved sleep, but the COSIT sleep goal correlated most strongly with improvement in the TFI auditory score. The correlations could be a result of cause and effect, or a third factor resulting in scores changing together. The direction of effects was not determinable in this study but would be a valuable outcome of further studies.

The COSIT identified 78 percent of participants as responders, and the TFI identified 43 percent as responders. These results suggest the TFI criteria are more conservative or, alternatively, the COSIT is more lenient or responsive to change. The COSIT criterion of “slightly better” appears to have high face validity as a minimum criterion for responding. Goal-specific therapy would be expected to improve COSIT scores more than the TFI because the individual therapy can be tailored toward addressing the person’s goals. The TFI is more generic than the COSIT and so may not fully capture goal-driven therapy benefit outside of general improvement in common tinnitus complaints. The nonuniform nature of the COSIT goals makes cross-individual and cross-population comparisons less easy than the TFI. The bespoke nature of the COSIT is both its strength and weakness. Our findings are counter to the opinion that the use of the COSIT is not feasible for clinical trials (Thompson et al., 2024). Our results suggest that the COSIT can be used as an effective outcome measure in trials and addresses gaps in understanding individual outcomes that closed-form questions cannot avoid.

Ideally, COSIT goals are “SMART,” that is, specific, measurable, achievable, realistic/relevant, and time bound (Doran, 1981). The goals set by our participants were largely consistent with previous studies using the COSIT (Searchfield, 2019) and the similar Goal Attainment Scale in Tinnitus (Wagenaar et al., 2024). “Focus” was the fourth most common goal in our analysis; an equivalent goal was not in the top five goals identified by Searchfield (2019; effect on concentration ranked eighth) and Wagenaar et al. (2024, less distracted was ranked 12th). This highlights the primary caveat to combining results from an open-set questionnaire: the coding and collapsing into themes may place participants’ statements containing several items into one or the other theme based on observer opinion or unintentional bias. The open-ended format also has the disadvantage of relying on clients’ recall. Thematic analysis does require judgment and will depend on the framing of participants’ responses and reviewers’ positionality.

The Kappa statistic indicated a fair agreement in responsiveness between the TFI and COSIT, but the questionnaires differed in the numbers divided into responders and nonresponders. In some cases, individuals had COSIT and TFI scores that were incongruent. One participant had a change in TFI of 50 points but was identified as a nonresponder by the COSIT. Six participants were responders according to TFI improvements of >13 but endorsed a COSIT category of worse or no change. These differences may be the result of some divergence in questionnaire meaning or represent the pre-post nature of the change in TFI compared with the posttherapy rating used by the COSIT. There may be a recency bias in the COSIT, distorting how bad the tinnitus was before therapy. Any error in completing the TFI might compound by assessing tinnitus twice, before and after therapy.

Goal-setting facilitates self-efficacy to focus attention on task-related solutions (Locke & Latham, 1990). In our clinic the COSIT plays an important role in initiating engagement between the client and clinician. The COSIT is completed through verbal questioning rather than a written task. It places the client in the center of the discussion and assists the clinician to move systematically through issues important for the client. It is client empowering by placing the client’s needs first and clinician enabling by facilitating the recommendation of realistic therapy solutions. Self-perceived inadequacy in tinnitus counseling is a barrier for audiologists to engage in tinnitus management (Searchfield et al., 2020). The COSIT is a very helpful tool for inexperienced clinicians because it facilitates engagement more naturally than closed questionnaires and provides a launch pad for developing a therapeutic alliance. Questionnaires are not without limitations and should be interpreted with care. Ultimately, how the person feels about their tinnitus should be the clinician’s primary concern, not the score on a questionnaire. We have had anecdotal reports that clients have been told by their clinician that a treatment has been successful because their questionnaire has shown improvement, even though the client does not think the tinnitus has changed. A process of continual improvement and sequential attainment of goals will lead to better outcomes than arguing a clinically meaningful degree of change has been achieved, somehow absolving the clinician from further therapeutic engagement.

Our recommendation for tinnitus clinical trials is that they include both a closed-set (TFI) and open-set (COSIT) evaluation of tinnitus impact on psychosocial factors. If tinnitus perception is an important consideration, a short-form rating of tinnitus magnitude should be useful (Schmidt et al., 2014). If psychological dimensions are of interest, these too can be evaluated (Husain et al., 2018). In the clinical environment the COSIT can be a time-efficient method to capture many of these aspects, including all that are a priority for the client.

CONCLUSIONS

The TFI and COSIT both differentiate responders and nonresponders with tinnitus intervention. The two questionnaires broadly measure similar constructs. Our results are consistent with the COSIT being an ideal “one for one” tool because of its high responsiveness to individual goals. The TFI is better suited as a “one for all” questionnaire because of its universal content. Clinical trials should consider use of both the COSIT and TFI because of their different, but complementary, strengths.

Abbreviations

AUC

area under the curve

CI

confidence interval

COSIT

Client Oriented Scale of Improvement in Tinnitus

ROC

receiver operator characteristics

SD

standard deviation

TFI

Tinnitus Functional Index

THI

Tinnitus Handicap Inventory

APPENDIX

Patient Characteristics

Characteristics	Numbers, score, percent
Sex, male/female	57/42
Age, y	57 (SD 13.8)
TFI total baseline	52.3 (SD 18.1)
Handedness	Right: 71, left: 19, both: 9
Family history of tinnitus complaints?	Yes: 35
How did you perceive the beginning?	Gradual: 64, sudden: 35
Was the initial onset of your tinnitus related to:	Loud blast: 21, whiplash: 1, hearing change: 23, stress: 12, head trauma: 3, other: 39
Does your tinnitus seem to PULSATE?	Yes: 10
Where do you perceive your tinnitus?	Right: 14; left: 9; both, worse right: 25; both, worse left, 11; inside head: 31; elsewhere: 9
How does your tinnitus manifest over time?	Intermittent: 11, constant: 88
Does the LOUDNESS of the tinnitus vary from day to day?	Yes: 62
Describe the LOUDNESS of the tinnitus using the slider scale from 1 (very faint) to 100 (very loud)	69.4 (SD 15.2)
Does your tinnitus sound more like a tone or more like noise?	Tone: 40, noise: 31, crickets: 23, other: 5
Please describe the PITCH of your tinnitus.	Very high: 19, high: 63, medium: 12, low: 4
What percent of your total awake time, OVER THE LAST MONTH, have you been aware of your tinnitus?	69.3% (SD 24.7)
What percent of your total awake time, OVER THE LAST MONTH, have you been annoyed, distressed, or irritated by your tinnitus?	49.4% (SD 26.4)
How many different treatments have you undergone because of your tinnitus?	None: 63, several: 20, many: 13
Is your tinnitus reduced by music or by certain types of environmental sounds such as the noise of a waterfall or the noise of running water when you are standing in the shower?	Yes: 62, no: 29
Does the presence of loud noise make your tinnitus worse?	Yes: 34, don’t know: 24
Does any head and neck movement (e.g., moving the jaw forward or clenching the teeth) or having your arms/hands or head touched affect your tinnitus?	Yes: 28
Does taking a nap during the day affect your tinnitus?	Worsens: 7, better: 7, no effect: 85
Is there any relationship between sleep at night and your tinnitus during the day?	Yes: 15, don't know: 51
Does stress influence your tinnitus?	Yes: 63, no: 36
Does medication have an effect on your tinnitus?	Yes: 8
Do you have a problem hearing?	Yes: 82
Do you have a problem tolerating sounds because they often seem much too loud? That is, do you often find sounds too loud or hurtful that other people around you find quite comfortable?	Never: 10, rarely: 24, sometimes: 38, usually: 20, always: 7
Do sounds cause you pain or physical discomfort?	Yes: 31
Do you suffer from headaches?	Yes: 32
Do you suffer from vertigo or dizziness?	Yes: 28
Do you suffer from any temporomandibular disorders?	Yes: 6
Do you suffer from neck pain?	Yes: 39
Do you suffer from other pain syndromes?	Yes: 19
Are you currently under treatment for psychiatric problems?	Yes: 7