PP605 - Prediction of Pure-Tone Thresholds by Speech Recognition Threshold in Elderly with Hearing Loss

            The procedure of pure-tone hearing evaluation requires subjective response and attention to listen unfamiliar sounds which is commonly difficult in elderly. 1226 patients recorded were randomly divided into training and testing sets. The speech recognition threshold (SRT) illustrated the most dominant variable in prediction pure-tone threshold (PTT) after adjusting with age. The PTT prediction formulas were constructed from the training set showed significantly high correlations between actual and predicted values in the testing set. It indicated highly explained variability of PTT at 0.5, 1, and 2 kHz. However, there was a noticeable high variation in PTT at 4 kHz-SRT discrepancy.

Summary:

Using of pure-tone thresholds (PTTs) for hearing aid fitting is most challenge to adjust hearing-aid gain in elderly due to variety of their physical and/or psychological factors against unfamiliar sound stimulus for interpreting pure-tone audiogram. The speech recognition threshold (SRT) uses a speech signal which provides more familiar and realistic to measure the patient's ability to recognize words. This study aimed to assess the discrepancy of SRT-PTTs and create the prediction formula of PTTs using SRT and adjusted with other potential factors. The finding formula will be a possible clinical tool to implement for more efficient hearing evaluation and hearing aid fitting in vulnerable elderly.

A retrospective study was retrieved from hospital information system in a tertiary hospital, Thailand. The data were obtained from those patients aged 60 - 90 years with diagnosis of hearing impairment enrolled during January 2011 to December 2021. All records of 1226 patients were randomly divided into training and testing sets using random sample function command in R. The data of all included participants were reviewed for the results of their PTTs, SRT, age, gender, hearing aid fitting, and related symptoms including cognitive loss, tinnitus, and movement disorders. Underlying diseases such as diabetes mellitus, cerebrovascular diseases, hypertensive diseases, disorders of lipoprotein metabolism, and depressive episode were also recorded. Multiple regression analysis was done to create the predictive relationship between the PTT and SRT with other variables, and performance checking by using the adjusted R-squared.

Moderately to very strong correlations between hearing levels measured by SRT and PTTs were found. Resultant multiple regression models of PTT prediction were formulated by the following variables based on SRT and age in training set. The overall prediction accuracies across the various frequencies of PTT prediction, a higher value of adjusted-R² indicated better performance which were significantly achievable by PTT prediction at 1 kHz (adjusted-R² = 0.8852) and 0.5 kHz (adjusted-R² = 0.8211) followed by 2 kHz (adjusted-R² = 0.741) and 4 kHz (adjusted-R² = 0.4109). According to the highest discrepancy of SRT and PTT at 4kHZ, the deviation of PTT at 4 kHz were influenced by age which led the inconsistency of hearing thresholds creating the variety of data set. A typical audiogram of geriatric population showed flat configuration at main speech frequencies and sharply sloping elevated hearing thresholds at frequencies above 2 kHz. Thus, PTT at 1 kHz showed the highest correlation and yielded the nearest agreement with SRT, followed by 0.5 and 2 kHz resulted in PTT-SRT discrepancies ranging from 0.261 – 4.347 dBHL.

The finding formulas of our study showed strong correlation of predicted PTTs and actual PTTs. The primarily practice to enhance a hearing level was a hearing aid fitting which required a PTT at specific frequency as a baseline for tuning the appropriate amplification of hearing aid. The prediction formulas should be used to apply for geriatric patients instead of pure-tone evaluation which may be took an extra time to repeat the hearing evaluation, especially among those who have complex health conditions.