Relationships Between Speech, Spatial and Qualities of Hearing Short Form SSQ12 Item Scores and their Use in Guiding Rehabilitation for Cochlear Implant Recipients

Abstract

Cochlear implantation successfully improves hearing in most adult recipients. However, in rare cases, post-implant rehabilitation is required to maximize benefit. The primary aim of this investigation was to test if self-reports by cochlear implant users indicate the need for post-implant rehabilitation. Listening performance was assessed with the Speech, Spatial and Qualities short-form SSQ12, which was self-administered via a web-based survey. Subjects included over 2000 adult bilateral or unilateral cochlear implant users with at least one year of experience. A novel application of regression tree analysis identified core SSQ12 items that serve as first steps in establishing a plan for further rehabilitation: items 1, 8, and 11 dealing with single-talker situations, loudness perception, and clarity, respectively. Further regression and classification tree analyses revealed that SSQ12 item scores were weakly related to age, degree of tinnitus, and use of bilateral versus unilateral implants. Conversely, SSQ12 scores were strongly associated with self-rated satisfaction and confidence in using their cochlear implant. The SSQ12 total scores did not vary significantly over 1–9 or more years’ experience. These findings suggest that the SSQ12 may be a useful tool to guide rehabilitation at any time after cochlear implantation. Identification of poor performance may have implications for timely management to improve the outcomes, through various techniques such as device fitting adjustments, counseling, active sound exposure, and training spatial hearing.

Introduction

The Speech, Spatial and Qualities of Hearing scale (SSQ) is a widely used hearing-specific scale that addresses perceptual listening challenges faced, in everyday circumstances, by individuals with any degree of hearing loss, whether listening in quiet or with competing sounds (Gatehouse & Noble, 2004). In the name of reducing administration time and questionnaire fatigue, short-form versions have been validated against the full, 49-item SSQ in cohorts of hearing aid (HA) users (Demeester et al., 2012; Moulin et al., 2019; Noble et al., 2012; von Gablenz et al., 2018) and cochlear implant (CI) recipients (Wyss et al., 2020). All versions have been utilized as a diagnostic tool and/or a means of describing real-world outcome benefits for HA and CI users (see review in Sanchez-Lopez et al., 2022).

The SSQ outcome is usually reported as a total score and may include three additional scores representing the three main subscales of speech understanding, spatial perception, and quality of speech sounds. A higher score represents fewer difficulties in coping with hearing loss (Demeester et al., 2012; Cañete et al., 2022; Sanchez-Lopez et al., 2022). The three-factor structure (speech, spatial, and quality), described by applying factor or principal component analysis (Akeroyd et al., 2014; Gatehouse & Akeroyd, 2006; Noble et al., 2012), has been questioned, however, given that there appears to be at least one other relevant subscale identified as “effort” or concentration (Akeroyd et al., 2014; Moulin et al., 2019). Gatehouse and Akeroyd (2006) described ten pragmatic attributes to consider more real-world experiences in grouping auditory experiences. These attributes formed the basis for selecting items included in the SSQ6 (Noble et al., 2012) and the SSQ12 (Noble et al., 2013). Although the authors’ intention was for the SSQ12 to be representative of the SSQ, it was not designed to mirror the subscales of the full SSQ; for example, it does not contain any questions associated with listening in quiet that might be more appropriate for CI users. The SSQ was constructed initially for patients with hearing impairment and conventional HAs as a central focus and not for recipients of CIs.

The purpose of the SSQ, cited by its authors, is to provide a means to support descriptions of hearing abilities and disabilities from the perspective of the listener. However, despite several mentions of its probable usefulness for indicating rehabilitative needs, the concept has received little attention (Banh et al., 2012; Noble et al., 2013, Sarant et al., 2020; Vannson et al., 2015; Warring et al., 2020). Recently, however, the short-form SSQ12 was shown to be a potentially useful tool beyond its diagnostic and outcome benefits purposes by Sanchez-Lopez et al. (2021). The authors revisited data from the original sample of subjects used to construct the SSQ (n = 1220 adults) and established 16 archetypes (patterns of responses) known as “benefit profiles” that could indicate rehabilitation paths for HA users. To develop these, they used auditory profiling to divide listeners into groups. This was accomplished by categorizing listeners according to audiometric findings and then comparing SSQ12 scores with outcome data from the Hearing Handicap Inventory (HHI) for adults. An example, for instance, might be high scores in the SSQ12 items but low scores on the HHI items would indicate that the rehabilitation approach should be directed more toward psychological support rather than primarily hearing support.

Taking a different approach, we reasoned that if associations between items were better understood, then it would be possible to represent SSQ12 items not in subgroups (i.e., to form subscales or auditory profiles) or in pragmatic attributes, but rather address each item separately and identify a set of core items that would support the SSQ12 as a means to indicate patient-specific rehabilitation needs. The goal in using this type of analysis, then, would be to develop a schema to aid in deciding the course of extended rehabilitation based on the self-reported performance of CI users.

In the current work, we revisited data gathered through an international survey that was first reported in a study associating tinnitus and SSQ12 ratings by CI recipients (Assouly et al., 2022). We observed that the study cohort was composed of CI users who presented highly variant item scores. The SSQ total and short-form scores for CI users have a wide range but are generally concentrated below 5.0 and can often be zero (Wyss et al., 2020). This contrasts with HA users and normal-hearing subjects who rarely have zero SSQ total scores (Moulin et al., 2019; Noble et al., 2013). Due to the non-normal distribution of single item scores in our sample, we chose to analyze SSQ12 responses using decision trees rather than more common regression type models. Classification and Regression Tree (CART) analysis is a nonparametric statistical method rarely used in CI research. To our knowledge, it has never been applied to any version of SSQ outcome data in CI recipients; in particular, it has not been used to investigate responses to and indicators of individual scores for SSQ12 items. The SSQ is not a unidimensional scale (e.g., Akeroyd et al., 2014; Moulin et al., 2019): each one of the standard three subscales have proven valuable in relating to different aspects of hearing (e.g., Dwyer et al., 2014; Moulin & Richard, 2016). Unfortunately, the SSQ12 was not constructed with those same subscales in mind (Noble et al., 2013) and, hence, require that we address outcomes on an item-per-item basis rather than a per-subscale basis.

The aims of this study were to extend the use of SSQ12, which could provide insight and useful feedback to clinicians and recipients to structure rehabilitation needs for adult CI users as indicated by their own self-report. To accomplish these goals, we analyzed SSQ12 data for a large cohort of survey responders by applying CART analysis to determine the strongest relationships between item scores and with a limited set of respondent variables to the SSQ12 items. We also compared survey ratings for categorical satisfaction and confidence to SSQ12 item scores, to identify a smaller number of core SSQ12 items.

Materials and Methods

Participants were recruited via a web-based survey system for patient-reporting that was designed by the company Cochlear Ltd. The initial primary aim of the survey was to obtain long-term post-market data on adverse effects and on self-reported performance (or hearing-related quality of life) using the SSQ12. The invitation to participate was sent directly to CI recipients via an email in which a link to the survey was provided. Once the survey had been received and the recipient voluntarily agreed to participate, instructions, and access to the survey were provided. The data collection was active from December 2019 until January 2020.

Data Source and Subjects

Survey invitees were adult CI-recipients who had received a CI500 series Nucleus® cochlear implant (Cochlear Ltd., Macquarie University, New South Wales, Australia). Survey invitations were sent to 7387 CI users in seven European countries (France, Germany, Ireland, Italy, Netherlands, Sweden, and the UK), were ≥18 years old at time of implantation, had at least one-year’ experience using their device and had received their implant(s) between 2011 and 2019 and thus used Nucleus CP810 or CP900 series sound processors. A total of 2322 agreed to complete the survey and share data. Respondents were from Germany (61.1%), France (14.2%), United Kingdom (8.5%), Sweden (6.8%), Italy (5.1%), the Netherlands (3.1%), and Ireland (1.2%).

An opt-out option (ethical waiver) was provided as part of the local legal and institutional requirements and those included in the study cohort provided an informed consent to take part in the survey and to share their data for later analyses. All data extracted from the web-based survey platform were anonymized.

Materials

The survey was in two sections. The first section consisted of the SSQ12 (Noble et al., 2013). As for all SSQ versions, responses to SSQ12 questions yield item scores on a visual analogue scale from 0 to 10. In this survey, subjects entered item scores in steps of 1 unit. Not applicable responses were also allowed. For ten of the SSQ12 questions, an item score of zero would indicate low self-reported ability in the situation (“not at all”) and ten would indicate complete competence (“perfectly”). For SSQ12 question 9 (Q9), zero would indicate that sounds are “jumbled” and ten that more than one sound can be heard at the same time and sounds are “not jumbled.” For Q12, about listening effort, zero would indicate a need to “concentrate hard” and ten “no need to concentrate.”

The second section of our survey was composed of additional questions (see Supplementary Material 1): There were questions related to the presence and degree of adverse effects (tinnitus, dizziness, ear pain or soreness, processor retention problems or discomfort, fatigue or tiredness, nonauditory sensations, and headache). If an adverse effect was present, then the subject was asked to rate the effect on a 5-point Likert scale from “Not at all bothersome” to “Extremely bothersome” (see Table 1 below). In addition, two questions asked for a 5-point Likert scale rating of perceived satisfaction with the device and confidence in using the CI (see Table 2 below). Other limited demographic details were collected from the subjects’ Cochlear CI registration records and included years of experience, unilateral or bilateral CI configuration, gender, and age at report.

Table 1.

Ratings for Additional Questions on the Severity of Reported Adverse Effects Reported on a 5-Point Scale.

Adverse effect	Not at all bothersome 1	Slightly bothersome 2	Moderately bothersome 3	Quite a bit bothersome 4	Extremely bothersome 5
Tinnitus (33%)	87 (4%)	241 (10%)	229 (10%)	167 (7%)	63 (3%)
Dizziness (18%)	24 (1%)	115 (5%)	130 (6%)	93 (4%)	48 (2%)
Ear pain (7%)	9 (<1%)	47 (2%)	53 (2%)	46 (2%)	9 (<1%)

Note: Other conditions reported with <0.1% of those surveyed are not shown here.

Table 2.

Ratings for Confidence and Satisfaction Ratings for Confidence and Satisfaction Reported on a 5-Point Scale.

	Very confident/satisfied	Confident/satisfied	Neutral	Not very confident/satisfied	Not confident/satisfied at all
Confidence	201 (9%)	1016 (44%)	712 (31%)	314 (14%)	79 (3%)
Satisfaction	937 (40%)	975 (42%)	291 (13%)	80 (3%)	39 (2%)

Subjects were asked “overall, how confident do you feel when interacting with others socially?” and “how satisfied are you with your cochlear implant(s)?”

All materials were translated using a certified translation process that included ensuring that the original meaning of the sentences was captured by being verified for cultural appropriateness by native speakers and in collaboration with one of the authors of the original English version and thereafter provided in the local language of the participant. The SSQ translation used in French was nearly identical to that used by Moulin et al. (2015) (they added a precision to Q8 about using sound only to judge direction). The Dutch translation was as used by Demeester et al. (2012), in German as in Kiessling et al. (2011), and in Swedish as in Stenbäck et al. (2023). We did not find a published validation of the Italian version.

Statistical Considerations

Variables collected in the study were summarized using mean and standard deviation for continuous variables and count with percentage for categorical variables. The Jonckheere–Terpstra test (Hollander et al., 2013) (a nonparametric one-way, analysis of variance, method suitable for data with an ordered predictor) was used to determine if there was a trend in SSQ12 total scores by categorical years of experience. If there was no evidence of a trend in SSQ total scores over years of experience, then a Kruskal–Wallis one-way analysis of variance was used to determine whether any pairs of times differed at the 5% level and then used to determine that the next phase of analysis could be conducted without consideration of the years of experience. Note that because of the smaller number of observations at some experience time points these were grouped as 1, 2, 3, 4, 5–8 and 9+ years for analysis. A previous analysis of SSQ12 total scores and tinnitus ratings and other respondent variables using linear modeling (regression) methods can be found in Assouly et al. (2022). However, in this study, we applied CART analysis (Breiman et al., 1984) to determine whether respondent variables and SSQ12 item scores were related, which SSQ12 items were associated with other SSQ12 items, and which SSQ12 items were most strongly associated with satisfaction or confidence. Such methods are useful because assumptions on the distribution of scores are not required; SSQ12 item scores and the ordinal satisfaction and confidence categories are not likely to have smooth bell-shaped distributions. Regression trees were applied where the outcome variables were SSQ12 item scores, and classification trees were applied to predict the ordinal Likert categories for satisfaction and confidence. The tree building process involves selecting the variables that branch the data into the purest response groups, using prespecified criteria such as minimum sample sizes and a cost associated with misclassification. This usually results in a large and complex tree and so, to avoid overfitting, each tree was pruned to create the simplest tree with the fewest branches, based on the best tree within one standard error of the minimum cross-validated error rate (Breiman et al., 1984). In addition, as an indicator of predictive strength, r² was calculated between predicted and actual values. Sensitivity and specificity were also calculated.

Satisfaction and confidence ratings were reduced to bivariate categories to simplify further analysis and interpretation. Responses of “very satisfied/confident” and “satisfied/confident” (see Table 2) were grouped against a neutral rating or “not very satisfied/confident” and “not at all satisfied/confident.”

Analyses were conducted using R version 4.1.0 statistical software (R Core Team, 2021). One SSQ12 unit, a clinically important value, was considered a relevant indicator of a difference between groups (Noble & Gatehouse, 2006). CART analysis was performed using the Rpart package within R (Therneau & Atkinson, 2019).

Results

The mean age of respondents was 57.9 years (range 18–95, SD = 15.7). Half of the respondents identified as female and 70% were unilaterally implanted. The rate of response to individual items of the SSQ12 was ≥99%; except item Q2 at 98%.

CI Experience

Respondents had a mixed range of experience from one year (18.5%) to two years (22.5%) with the majority greater than three years (59.0%) (see Figure 1). There were generally fewer respondents with experience between six and eight years. This was attributed to the method of recruitment to survey groups that were implanted approximately five years before the survey. The slight increase in numbers for nine years and greater may reflect additional recruitment after sound processor renewal.

Figure 1.

SSQ12 total scores grouped by years of experience for 2322 CI users. Large black points are means, middle lines are medians and boxes show 25th and 75th percentiles. Whole years are represented where “1” year indicates 1–2 years of use, “2” for 2–3 years of use, etc. Years 5–8 and 9+ were grouped because of small numbers of participants. Numbers at the top of each column are the number of participants in that group. Red dots indicate total scores where item scores were <6 for Q1, <5 for Q8, or <6 for Q11, blue dots the remainder (see low vs. high core item scores section).

Overall, SSQ12 total scores were stable over time (Figure 1) with no evidence of a significant trend by years of experience (Jonckheere–Terpstra p = .85) or between any pairs of experience time points (Kruskal–Wallis p = .08). A very large range of SSQ12 total scores was observed across all time intervals. Therefore, for the following analyses of SSQ12 item scores, we combined the data across years of experience.

Adverse Effects, Satisfaction, and Confidence

Respondents reported the presence of tinnitus (33%), dizziness (18%), and ear pain or soreness around the implant site (7%) (Table 1). Twenty percent of respondents had tinnitus rated “moderately bothersome” or more severe. Dizziness was reported as “moderately bothersome” or more severe in 12% of respondents. Ear pain was described as “moderately bothersome” or more severe in 4.6% of the total study group. Other adverse effects, potentially associated with a CI and reported by more than 0.1% of respondents, were sound processor retention problems or discomfort each at 1.7%; fatigue or tiredness each at 1.2%; nonauditory sensations at 0.6%; and headache or general pain each at 0.4%.

Ratings for confidence in social communication and satisfaction with CI are given in Table 2. Just over half of respondents (53%) indicated they were “very confident” or “confident” and 17% “not very confident” or “not confident at all.” More than 80% of respondents indicated they were “very satisfied” or “satisfied” and 5% “not very satisfied” or “not satisfied at all.”

Item Scores

The scores given for each SSQ12 item are summarized in Figure 2. Mean and median scores tended to be around 4–6 for most items except Q11 at seven. All items except Q1 and Q3 had a substantial number of extreme (0 or 10) scores whereby circles over the shaded 0- or 10-score areas indicate at least 10% of the responses to that item. Notably, Q2 had a greater number of zero scores compared with scores of 5. Similarly, Q10 had an approximately equal number of scores of 0 or 5. Interquartile ranges spanned four to five score values, with the large numbers of zero scores, mentioned above, producing larger first quartile to median ranges than median to upper quartile ranges in five of the 12 items.

Figure 2.

The number of discrete item scores 0–10 per SSQ12 item are indicated by the size and darkness of the shading. Black points indicate the means, middle lines the medians, and the error bars the 25th and 75th percentiles. Open circles indicate more than 10% of the cohort gave an item score at either 0 or 10. Boxes summarize the main finding of the regression tree analysis between SSQ12 items and confidence and satisfaction ratings.

Classification and Regression Tree Analysis

Respondent Variables and SSQ12 Item Scores

The regression trees in this subsection were constructed for each SSQ12 item using the respondent variables of age, gender, unilateral or bilateral implant, and bothersomeness of tinnitus and dizziness. An example of a regression tree is given in Figure 3 for SSQ12 Q1. Each ellipse in the tree shows the mean Q1 score and percentage of respondents in that branch of the tree. In this tree, two variables were found to be the best predictors of Q1 item scores: age and tinnitus. Respondents with age ≥ 76 years (13% of the cohort; the left most branch of the tree) had a predicted mean Q1 score of 4.2. Those of younger age (in the right-hand side of the tree) further branched into two groups. These two groups were based on report of tinnitus as moderately bothersome or worse versus slightly or not at all bothersome. Those with moderately bothersome or worse tinnitus had a predicted mean Q1 score of 4.7, while those with slightly or not at all bothersome tinnitus had a predicted mean Q1 score of 6.

Figure 3.

Regression tree for respondent variables associated with Q1. Numbers in boxes are the mean item scores for branches/leaves, and the percentage of subjects represented in the branches/leaves.

Across all SSQ12 items, the respondent variables most frequently associated with SSQ12 item scores were tinnitus (SSQ Q1–Q5, Q9), age (SSQ Q1, Q2, Q4, Q10), and unilateral vs bilateral CI (SSQ Q6, Q8, Q11), in that order. For example, respondents generally gave lower item scores for SSQ12 items when “moderately bothersome” or a more severe category was reported for tinnitus. This cut-point was the same for all seven tinnitus-associated questions. Both tinnitus and advanced age were associated with scores for Q1 (≥76 years) and Q4 (≥71 years). An age cut alone was seen for Q10 (≥63 years). Bilateral CI typically gave higher scores than unilateral CI for Q6, Q8, and Q11. No respondent factors were found to significantly associate with scores for Q7.

Differences in mean item scores for each branch or the extreme leaves indicate the strength of relationships. This difference was 1.8 item scores for Q1 (6 vs. 4.2) using both the age and the tinnitus branches: this was the largest difference but was still less than half of Q1's interquartile range (Figure 2). Differences branching on respondent variables for the remaining items were less at 0.8 (Q10), 1.0 (Q2, Q5, Q8, Q11), 1.1 (Q3, Q9), and 1.4 (Q4, Q6, Q12), and generally spanned only one quarter of the interquartile ranges. Thus, although some respondent variables did significantly explain variations in item scores, they did not serve as strong discriminators of item scores. Using the predicted item scores (i.e., means of end leaves) versus actual item scores, the variance explained was relatively low, ranging from 2% (Q2, Q5, Q8, Q9, Q10) to a maximum of 6% for Q1.

Relationships Between SSQ12 Items

Spearman correlation coefficients between item scores are given in Figure 4 (left/lower triangle). Coefficients were all positive and those ≥0.55 are indicated in bold. The usual main subscale groupings are indicated by boxes. Item scores for all questions within the speech and spatial subscales had moderate to strong positive correlations between 0.6 and 0.8, showing strong relationships between the original subscale questions. However, item scores within the Qualities subscale had similar correlations to item scores in other subscales, potentially indicating a more tenuous subscale. This pattern has been observed in previous data from normal-hearing or HA users (Gatehouse & Akeroyd, 2006; Moulin et al., 2019). Some of the higher correlations are also evident in the scatterplots shown in Figure 4 (upper/right triangle; see Q3 and Q4; Q7 and Q8). However, some of these patterns reflect a conditional relationship. Clear examples of conditional relationships are between Q1 and Q2, and Q11 and Q10 where points are concentrated on and to only one side of the line of equality. This indicates that scores for Q2 were most often the same or less than for Q1 and similarly for Q10 that was most often the same or smaller than for Q11. Less clear but similar patterns were observed for Q3, Q4 and Q5 versus Q1, and perhaps Q4–Q9 versus Q11. These can be contrasted with the apparent concordance between Q7 and Q8, where points are concentrated on the line and to both sides of equality (also to a lesser extent Q7 and Q6).

Figure 4.

Lower left diagonal: Spearman correlation coefficients between scores for different items. Coefficients ≥0.55 in bold. Upper right diagonal: Scatter plots with more intense color indicate more points for a given intersection of item scores are shown. The small top and right axes numbers indicate the item scores 0–10.

The conditional relationships seen in our dataset, as mentioned above, imply that some situations are easier than others and that a higher item score for an easier situation or perception is not necessarily associated with a higher item score for a more complex situation or perception. Thus, these findings potentially guide the clinician to choose a more focused approach when structuring rehabilitation, since the items themselves indicate a hierarchy of difficulties encountered. These findings reflect the necessity of acquiring a certain level of skill in a simpler situation, to advance to more complex skills. To identify the strongest relationships between SSQ12 items, the regression tree technique was applied. An example of a regression tree for the prediction of item scores on Q12 is shown in Figure 5 (Regression trees pertaining to each item are provided in Supplementary Material 2). The overall cohort had a mean item score of 5.3 for Q12. The first branch suggests that Q11 yields the best prediction of item scores for Q12, followed by further branches based on Q4 (lower mean item score) and Q5 (higher mean item score) as well as branches based on item scores again for Q11 and Q1. After pruning, most items yielded 3 branch levels, except for Q9 and Q10 that had one further branching.

Figure 5.

Regression tree for other SSQ12 items that predict Q12 scores. Circle sizes indicate levels of interrelationship according to branch level. T = True; F = False.

SSQ12 Constellations

Figure 6 summarizes the regression tree analyses for all SSQ12 question-to-question relationships and the respondent factors. The vertical axis denotes the predictor question and the horizontal axis the dependent question for which the item score is being predicted. The strength of the relationship is denoted by circles “O,” where a larger circle represents a stronger relationship according to branch level (see example in Figure 5). Four notional constellations were identified: loudness, clarity of sounds, and speech in single- and multitalker backgrounds. These are labeled as loudness, clarity, single (talker), and multi (talker). These constellation names were given to differentiate them specifically for CI users from the pragmatic attributes and main subscale names used in studies of normal-hearing listeners and HA users. We did not include Q12 in the constellations as the responses may be considered the sum total of a number of factors that may include Q11 (sound clarity) and Q4 and Q5 (multitalker background).

Figure 6.

Summary of trees for respondent factors and SSQ12 items. The original SSQ subscales and the new constellations (“Single” – single talker, “multitalker” – multiple talkers, loudness, and clarity) described in this study are given on the vertical and horizontal margins, respectively. Colors on the diagonal indicate significant relations for respondent variables and SSQ12 items with numbers indicate branching by age (years). Sizes of circles indicate the level of interrelationship between SSQ12 items.

Item scores for Q1 and Q2 (single-talker background) strongly corresponded to the similarity of the situations described, with Q1 having a conditional effect on the more difficult task in Q2 that, in addition, involves switching attention (see also Figure 4, top left-most scatter plot).

Similarly, the situations and tasks described (i.e., multitalker background) are somewhat similar for Q3, Q4, and Q5. Clarity (Q9, Q10, and Q11) appears to be driven by item scores for Q11 about the clarity of single sounds, as a prerequisite for situations and tasks that involve separating multiple sounds. Finally, we have considered “loudness” rather than “distance and movement” and “localization” to describe the last constellation (Q6, Q7, and Q8). We propose loudness perception as an underlying factor for this constellation because although item scores for Q6 and Q8 were significantly associated with bilateral versus unilateral implantation, the actual difference in mean item scores was relatively small, as was the variance explained (4% and 2%). We envisage that for the described tasks, two ears are not necessary (see loudness discussion below). For example, monaural loudness cues can be used in the situations described in Q7 and Q8, as can monaural cues combined with head turning for Q6.

The four items with the highest mean scores were Q1, Q8, Q11, and Q12. Q1, Q8, and Q11, we define as core items as they were the easiest, being most often given higher scores than the others. We discuss this concept in more depth below. As mentioned above, Q12 appears to represent an outcome rather than an ability. One can argue that “effort” is required where the perceptual ability is poor or the situation difficult, rather than a skill in its own right. Effort is the work involved; whereas the skills involved in listening with a CI are mainly learned and may also depend on innate abilities.

SSQ12 Items Related to Satisfaction and Confidence Ratings

Classification tree analysis identified Q11 as the best predictor of satisfaction, with a branch at a score of 5. There was a second branch based on Q4 for those with Q11 scores <5 (Figure 7, upper). This classification tree resulted in a sensitivity of 0.33 to detect low satisfaction, with a specificity of 0.97. The sensitivity was improved to 0.50 using only the single branch for Q11 score <5, with a reduced specificity of 0.90.

Figure 7.

Classification trees relating SSQ12 item scores to satisfaction and confidence ratings. We note that using the first-level classifications alone resulted in higher sensitivity for “neutral or not satisfied/confident,” compared with complete trees.

For confidence in communication, classification tree analysis identified Q3 as the best predictor, with a branch at a score of 5. There were further branches for Q12 at 6 and Q5 at 4 (Figure 7, lower). This classification tree resulted in a sensitivity of 0.63 to detect low confidence, with specificity of 0.83. However, a higher sensitivity of 0.69 was obtained with just the single Q3 branch at a score of 5, at the cost of a lower specificity of 0.75. Thus, the simpler, single Q3 branch appears sufficient and more basic as a rule of thumb to predict low confidence in communication and has face validity based on the situation described. Q3 is included in the 15iSSQ (originally 1.11) and thus consideration of the factors of confidence and satisfaction would also be available for the 15iSSQ.

Definition of Core Items and Recommendations for Clinical Use

We reason that core skills are fundamental and underpin more complex skills; that is, where those skills indicated in Q1 and Q11 were rated high, other items can be rated high according to the conditional relationships within constellations cited above. We add that Q8 was largely indicative of performance for Q7 and Q6 (Figure 4, Figure 6). Cochlear implant users who scored low on core items scored low across 10–12 items; whereas those who scored high on all core items (Q1 > 5, Q8 > 4, and Q11 > 5) scored consistently high across 10–12 items resulting in generally high SSQ12 total scores (Figure 1, red and blue dots, respectively). The implication is that rehabilitation directed at enhancing core skills may eventually lead to improvements in more complex skills.

Q1 (Talking with One Person with TV on)

Q1 was rated very high in the constellation of single/multitalker. Q1 was second only to Q11 and is thus one of the easier listening tasks. This suggests that in terms of understanding speech, if CI users do not do well in the simplest situation as described in Q1 (reporting an item score <6), then they are less likely to do well in the more complex situations described in Q2–Q5. This may have a knock-on effect in that they are not confident in communication as indicated by Q3. Note that SSQ12 item Q1 corresponds to 15iSSQ item 1 (Moulin et al., 2019) and SSQ49 item 1.1 (Gatehouse & Noble, 2004).

Q8 (Vehicle Approaching/Receding)

Among those items in the constellation associated with loudness, Q8 represented the easiest task. Tree analysis showed a weak effect of unilateral versus bilateral CI, where bilateral users reported higher item scores. Along with Q7, Q8 also predicted item scores for Q9 and Q10. It is our position that many issues linked with clarity and understanding may be related to coding intensity, that is, loudness perceptions. Q8 was associated with four different SSQ12 items: Q6, Q7, Q9 and, to a lesser extent, Q11. Question 8 was not selected for inclusion in the 15iSSQ (Moulin et al., 2019). We suggest using 15iSSQ item 7 (i.e., Q6 here) in place of SSQ12 item Q8 for the loudness constellation. For those using the SSQ49, Q8 is item 2.13.

Q11 (Clarity of Everyday Sounds)

Within the constellation associated with clarity, Q11 received the highest item scores. Categorical tree analysis showed it to be associated with satisfaction. We suggest that item scores for Q11 < 5 can be used as a guide for dissatisfaction. Separation of sounds (Q9) and classification of musical instruments (Q10) appears to be driven by item scores for Q11 about the clarity of single sounds as a prerequisite for situations and tasks involving multiple sounds. Q11 also predicted scores for Q12; that is, higher Q11 scores generally indicate less effort required in listening situations, but with some evidence of a conditional relationship (Figure 4, bottom right). Note that SSQ12 item Q11 here corresponds to 15iSSQ item 15 and SSQ49 item 3.9.

Low Versus High Core Item Scores and Total Scores

We observed that item-to-item root branches gave greater than two SSQ12 points differences in means; further, that final tree leaves spanned interquartile ranges for all items. Core items were often found as item-to-item predictors by the CART analysis, either at root or other branches (Figure 6). In addition, the conditional relationships found for Q1 and Q11 indicated that these items weighed-in heavily on SSQ12 total scores. Looking through the item-to-item trees, we found branches on Q1 and Q11 scores at scores of 6, and for Q8 at a score of 5. These mostly or nearly correspond to median scores for the core items (Figure 2). Using these core item scores, we divided the population into two groups: those with Q1 or Q11 scores <6, or Q8 scores <5 (Figure 1, red dots) and those with higher scores on all core items (Figure 1, blue dots). We observed that these core item score classifications approximately divided the population into those with SSQ12 total scores below and above 6, or at the 75th percentile. A total score of 6 could be considered satisfactory performance from a clinical point of view.

Recommendations for Clinical Use

The following outlines simple steps to interpret the SSQ12 total score, the core item scores and the noncore item scores. If observing that the total SSQ12 is <6, proceed to step 1.

• Step 1: Observe if there are low scores on any of the Core-Questions, < 6 on Q1 and Q11, and <5 on Q8: If yes, then timely post-implant rehabilitation is recommended relating to speech understanding activities and programming options that may influence clarity and loudness, respectively.

Then,

• Step 2: Consider the conditional relationships within constellations; that is, is Q2 much less than Q1, Q6, and Q7 much less than Q8, or Q9 and Q10 much less than Q11? If yes, then consider working on speech in one or more competing-talker conditions, and movement and location rather than just distance, and identification of sounds versus discrimination, respectively.

Discussion

The primary aim of this study was to identify details by which patient-directed, post-implant rehabilitation needs could be indicated as an outcome of self-assessments made possible by interpreting SSQ12 item scores. The survey avoided imposing time commitments on already overburdened clinics in collecting data through a rapid and efficient means of building a large data pool. We reframed the interpretations of the SSQ12 by analyzing the relationships between items rather than total or subscale scores. To this end, decision tree analysis was applied to determine the strongest relationships. These, then, could be used to help understand what differentiates poor and good performers (low vs. high item scores). Thus, we propose a set of recommendations (see above) to guide clinicians in quickly scanning SSQ12 item scores to distinguish those CI users who may be in need of more immediate post-implant rehabilitation verses those who function well with spontaneous learning and require minimal follow up.

The findings of this study should be interpreted as a guide rather than a definitive instruction for applying post-implant rehabilitation. They are intended to empower clinicians in making timely decisions to encourage and enhance more effective use of CIs and reduce in-clinic time.

Implications for Rehabilitation

It is beyond the scope of this paper to describe specific rehabilitation approaches. Each clinician will wish to address rehabilitation options using their own best practices, experience, and resources. Time constraints, manpower, need for clinical presence, scheduling, caseload, cost, and more will factor into the decision to offer various in-clinic rehabilitation options to CI users (Zhou et al., 2013). A satisfied and confident user will require less support, allowing for only biannual or annual follow-up. The SSQ12 total scores combined with core item scores may support this utilitarian purpose.

It should be noted, however, that those reporting themselves as confident in communication situations were more likely to be satisfied with the implantation and provide higher SSQ12 scores (data not shown). Clearly, satisfaction, and confidence are independent variables; for instance, it is possible for an individual to rate high satisfaction and low confidence, while more unlikely to report low satisfaction and high confidence.

The ability of a specific CI user to take advantage of a particular rehabilitation strategy, if at all, may be indicated by changes in core item scores. It is yet to be observed if scores on these core items will be the first to improve following rehabilitation treatments or whether additional factors (predictors) will contribute to positive change. A total of 68% of subjects in this sample had SSQ12 total scores <6, and 50% scored <6 on item Q1, 42% scored <5 on item Q8, and 29% scored <6 on item Q11. This pattern of scores would indicate the need to apply rehabilitation methods for single-talker speech understanding more than for loudness and clarity. We note, however, that these findings were gathered from a survey that took place when subjects had already experience at least one year of CI use.

As mentioned earlier, a previous analysis of SSQ12 total scores and tinnitus ratings for the current data set can be found in Assouly et al. (2022). In our more extensive analysis, we identified six questions where item scores were associated with tinnitus severity, along with four associated with age and four with bilateral versus unilateral implantation. The addition of a second implant could improve subjective performance; however, we point out that the impact on item scores was relatively small. Thus, other interventions such as training and device programming may produce greater improvements.

There are several good systems for applying performance profiles, such as that developed for HA users by Sanchez-Lopez and colleagues (2021). Unfortunately, the Sanchez-Lopez profiles rely heavily on audiometric threshold data but, as pointed out by Mertens et al. (2016), CI and HA users do not have similar hearing profiles. Furthermore, it is our contention that CIs are vastly different from HAs although rehabilitation approaches have typically been similar. With a HA, the basic functions are mainly to increase amplitude and provide frequency shaping of acoustic signals. On the other hand, an alternative auditory platform is introduced to the brain for interpretation via electrical input through a CI. Thus, the CI user's needs may be quite different from HA users and rehabilitation options may also be different. Further, there is the challenge in providing rehabilitation to bimodal users where it may be beneficial to determine which device, CI or HA, dominates in active listening conditions.

Traditionally, first-line approaches for CI users who may need rehabilitation are device programming and counseling.

Cochlear Implant Device Fitting and Loudness

Fine tuning a CI is an important adjunct to the rehabilitation process and, consequently, outside of standard fitting procedures. Additional adjustments may be indicated, particularly associated with intensity/loudness judgments (e.g., noise canceling properties, loudness steps, compression algorithms, and threshold-to-comfort level electrical dynamic ranges). Although interesting, these are perhaps far away from practicality in standard programming of CIs (see McBeath & Neuhoff, 2002 and McCracken & Neuhoff, 2021 for discussion on loudness vs. pitch perception).

Sound clarity (Q11) may be improved by reducing current levels in either high-frequency or low-frequency channels, or even deactivating “screechy” high-frequency channels or “boomy” low-frequency channels. The effect of frequency-to-electrode allocation on speech recognition (Q1) has been explored in the past (e.g., Başkent & Shannon, 2005). Similarly, speech training has been explored and shown to improve sentence recognition in CI users (Plant et al., 2015).

There is a large body of evidence to support CI programming adjustments to features associated with intensity and loudness perception. It has been reported that CI users do not differentiate degrees of intensity as well as those with acoustic hearing (Nikakhlagh et al., 2015). We contend that Q6–Q8 are best related to coding intensity. Item Q6 is about localizing a barking dog, whereas Q7 and Q8 are about distance and movement. Binaural hearing may not be necessary to achieve these spatial tasks, although we found bilateral CI users had slightly higher average item scores than unilateral CI users for Q6, Q8, and Q11. Bilateral CI users rely mainly on interaural level differences (ILDs) to localize sounds (e.g., Klingel & Laback, 2022), so that binaural loudness perception is important. However, monaural loudness cues could be used in the situations described in Q7 and Q8, as could monaural cues combined with head turning for Q6, since the real-world stimuli would typically be repeated or of longer duration. Thus, we propose to give emphasis to loudness adjustments as they may directly impact scores for SSQ12 items Q6–Q10, and possibly indirectly influence Q3 and Q11. Sound processor preprocessing may have some effect on loudness perception. For example, deactivating or reactivating input compression or other automatic gain systems may improve loudness perception (e.g., Spahr et al., 2007).

Indeed, for bilateral users, the compression functions used in CI sound coding may be a factor in sound localization. Potts et al. (2019) suggest that ILD perception can be trained but that interaural time differences (ITD) are more difficult. They suggest that bilateral CI listeners primarily use ILD cues rather than ITD cues for localization. But, there is also evidence that training can improve ITD perception in CI users (Nikakhlagh et al., 2015).

Interestingly, establishing symmetry between ears (whether bilateral CI or bimodal) may be beneficial. Studies have shown that when two ears tested independently had equivalent performance on speech tests; there was a bilateral benefit, especially in noise (Yoon et al., 2011). A significant binaural advantage in quiet has also been demonstrated by Mosnier et al. (2009) in their group of CI users who demonstrated symmetrical configurations between ears compared with those with asymmetrical configurations. Therefore, in the case of bimodal or bilateral users, loudness balance and loudness growth should be investigated (Blamey et al., 2000). This could also be the case for single-sided deafness (SSD) CI users (Buss et al., 2018). Further, we speculate that loudness perception may play a role in identifying sounds (Q9) or distinguishing musical instruments (Q10). Perhaps improved adjustments for intensity parameters will improve scores for those two items.

Direct Counseling

The focus of direct counseling may well be motivation and establishing appropriate expectations. Concepts related to informational counseling (Boothroyd, 2007; Kelly et al., 2013) may include general device use, how to manually change programs, increasing active listening time (Buss et al., 2018; Razorenova et al., 2020) and exposure to unique and varying acoustic environments ranging from easy to difficult, and how to make use of digital platforms as self-training methods.

The overall aim is to increase use of the CI to improve listening skills and deter social isolation that may be due to hearing loss. Data logging information could be useful to this end as supported by investigations that suggest improvements occur with consistent use of a CI (see Holder & Gifford, 2021).

Follow-Up Use of the SSQ12

Repeated use of the SSQ12 offers a viable means of tracking progress; however, it is recommended not to repeat administrations too frequently: perhaps at the one-month follow up and then a pause of several months. The focus for readministrations, other than scheduled annual follow-ups, would be particularly appropriate for those with listening issues that appear in many domains. Finally, limit discussing the SSQ12 findings with the CI user, thus avoiding self-rating bias, but rather inform them about the outcomes of their self-assessment as it relates to uncovering needs for a rehabilitation plan or to demonstrate that the outcomes have supported their current path of rehabilitation training.

Limitations of This Study

A selection bias may be present, as only those who were technologically capable of using Internet and who also voluntarily agreed to participate in the survey were respondents. Further, only follow-up data were requested for SSQ12 and there would have been a natural attrition rate not represented in the data, for example, not all CI implantees during the period of interest took part in the survey.

It may be considered a shortcoming that the survey included only information gathered postoperatively, providing no comparison to preimplant scores. Two relevant points can be made in support of this study, however: first, one must consider recall bias although retrospective studies are often part of hearing science research (Andersson et al., 2021; Glen, 2022) and, second, participants were queried one time without involvement of clinic personnel, which is considered an advantage.

Future Research

There is much more to investigate in understanding perceptual skills as they pertain to the SSQ12. The use of CART analysis is a new concept applied to CI research that raises a wide range of unanswered implications requiring further studies. For instance, it would be interesting to apply CART analysis to a group of pre-post rehabilitation recipients to help explain therapeutic effects. The current study only included subjects with post-implant data and we did not have information on the use of contralateral HAs (bimodal) or which ear was implanted. It would be useful to compare SSQ12 item scores in a similar way for bilateral CI and bimodal respondents. We note that the bimodal option was not available in our respondent variables data, but that it is assumed that 50–60% of users may have retained use of a HA in the nonimplanted ear (Devocht et al., 2020). It may be particularly interesting to consider confidence and satisfaction implications in those groups. For example, comparing SSQ12 item scores, before and after providing significant adjustments to programming features relating to loudness perception would be useful to estimate the sensitivity of item scores. Another question to consider with respect to core items is how much of a particular item (skill) can be converted into another skill or how much improvement in scores for core items will affect other items? This can only be tested in light of pre-post rehabilitation evaluations.

Conclusions

The current SSQ12 data indicate that self-reported hearing performance in adult CI recipients was stable over the long term; from at least one year of experience up until greater than nine years of experience. This large data set also allowed us to investigate the structure in scores between the SSQ12 items. Self-rated satisfaction with the device was most strongly associated with SSQ12 item Q11, relating to sound clarity, and confidence was most strongly associated with item Q3, relating to speech communication in a multitalker situation. We identified three core items, Q1 (single talker), Q8 (loudness), and Q11 (clarity/satisfaction) with the highest scores within their subscales, such that scores for the other items tend to correlate and/or be conditional on scores for the core items. The core items provide solid signs as to whether a particular individual is perceiving significant difficulty in utilizing hearing through a CI when in various everyday circumstances. The scores for core items and the conditional relationships for items associated with them guide further investigations that may provide the most efficacious follow-up for CI users who have reached a plateau in benefit (i.e., after one year) or present as being dissatisfied or showing low confidence when listening in communication situations. A simple set of recommendations has been provided for clinical use.