Discrimination loss vs. SRT: A model-based approach towards harmonizing speech test interpretations

TL;DR

The proposed approach can be used to assess additional differences between speech tests and its properties relate to data availability for individual patients, and all SRT procedures are influenced by the uncertainty of the word recognition scores.

Abstract

Objective: Speech tests aim to estimate discrimination loss or speech recognition threshold (SRT). This paper investigates the potential to estimate SRTs from clinical data that target at characterizing the discrimination loss. Knowledge about the relationship between the speech test outcome variables--conceptually linked via the psychometric function--is important towards integration of data from different databases. Design: Depending on the available data, different SRT estimation procedures were compared and evaluated. A novel, model-based SRT estimation procedure was proposed that deals with incomplete patient data. Interpretations of supra-threshold deficits were assessed for the two interpretation modes. Study sample: Data for 27009 patients with Freiburg monosyllabic speech test (FMST) and audiogram (AG) results from the same day were included in the retrospective analysis. Results: The model-based SRT estimation procedure provided accurate SRTs, but with large deviations in the estimated slope. Supra-threshold hearing loss components differed between the two interpretation modes. Conclusions: The model-based procedure can be used for SRT estimation, and its properties relate to data availability for individual patients. All SRT procedures are influenced by the uncertainty of the word recognition scores. In the future, the proposed approach can be used to assess additional differences between speech tests.

Figures (8)

• Figure 1: (A) Data flowchart, overview about data cleaning and data availability for different $SRT$ estimation methods. Note that, after excluding $WRS_{max}$, not all remaining data points necessarily fall within the defined slope area, meaning the actual number of data points in the slope area can be less than $N_{DP}-1$. The colored boxes represent the total number of patients included in the final analysis, as shown in Figure \ref{['fig:maxSI_pta']}. Examples of data availability and resulting fit are depicted in (B) for the empirical slope-based, (C) for the SII-slope-based, and (D) for the normal-hearing slope-based $SRT$ estimation procedure, respectively. Continuous lines in subfigures (B) and (C) indicate the slope area. For illustrative purposes, Plomp's $A$ and $D$ component are indicated in subfigure (B), and the discrimination loss $100\%-WRS_{max}$ is visualized in subfigure (D). Abbreviations: Audiogram (AG), Freiburg monosyllabic speech test (FMST), speech recognition threshold (SRT), pure-tone average (PTA), maximum word recognition score ($WRS_{max}$), speech intelligibility index (SII).
• Figure 2: Clinical data availability in the $WRS_{max}$interpretation mode. (A) Number of patients for different numbers of measured FMST test lists. The data availability according to Figure \ref{['fig:data_flowchart']} is color-coded. (B) Scatterplot of existing combinations of $WRS_{max}$ and $PTA$ in the data set. Colors correspond to the same groups as in (A), the marker size indicates the number of patients in each combination, in logarithmic scaling.
• Figure 3: $SRT$ (depicted as $SRT$ loss, the difference to normal-hearing reference $SRT_{NH}$) over $PTA$ for (A) the empirical slope estimation, (B) the SII-slope estimation, and (C) the psychometric function fit with normal-hearing slope and $WRS_{max}=100\%$. The marker size indicates the number of patients in logarithmic scaling. The grey line represents the diagonal where $SRT$ corresponds to $PTA$.
• Figure 4: Comparison of empirical slope and SII-slope-based estimation procedure. (A) $SRT$, (B) slope, and (C) $D$ component. The marker size indicates the number of patients in logarithmic scaling. Violet lines represent percentiles of the respective SII-slope-based variable for given empirical variable ranges, the thicker line represents the median.
• Figure 5: Predicted vs. observed $SRT$ difference ($SRT_f - SRT_h$) for fully-determined patients. The marker size indicates the number of patients in logarithmic scaling. Lighter colored lines represent the $10^{th}$, $30^{th}$, $50^{th}$, $70^{th}$, and $90^{th}$ percentiles of the respective predicted $SRT$ difference for given observed $SRT$ differences, the thickest line represents the median.