Automated Tool Developed at Mass Eye and Ear Distinguishes Individuals With Tinnitus From Fraudulent Claims

The diagnosis of tinnitus currently relies on self-reported measures, and, like pain, it is sometimes fraudulently reported in claims for disability benefits, release from military service or early retirement.

Subscribe to the latest updates from Otolaryngology Advances in Motion

Error: Please enter a valid email address.

Email Address Submit

Daniel B. Polley, PhD, who holds the Amelia Peabody chair in Otolaryngology–Head and Neck Surgery at Mass Eye and Ear and directs the Lauer Tinnitus Research Center there, and colleagues have developed a new automated, quantitative system for distinguishing between legitimate cases of tinnitus and fraudulent cases.

The system should also prove useful for assessing the characteristics and severity of tinnitus, as the team says in npj Digital Medicine.

Methods

The researchers compared two groups:

• Tinnitus group—24 patients, average age 52, recruited from Mass Eye and Ear who had reported bothersome, chronic tinnitus for at least one year
• "Malingerers"—28 individuals, average age 54, who affirmed they did not have tinnitus and were trained on what tinnitus sounds like

Using tablet computers and calibrated headphones, the participants performed three self-directed psychoacoustic tasks at home. The malingerers were asked to perform all tasks as if they heard a constant phantom sound:

• Visual analog scale (VAS)—Subjects used a virtual slider on the tablet computer screen to indicate how loud their tinnitus was that day
• Minimum masking level—Subjects adjusted a virtual slider to the lowest noise level that masked their tinnitus
• Tinnitus matching—Subjects used virtual sliders to match the pitch, loudness, bandwidth and modulation of their tinnitus

Subjects performed five test sessions over a two-week period, and during each session they performed multiple repetitions of each of the three tasks

Classification Based on Final Value of Slider

Logistic regression models were able to distinguish patients from malingerers based on the final value of each slider. Combining all slider values (one for the VAS task, one for the masking level and four for matching), accuracy was 68% and the area under the receiver operating characteristic curve (AUC) was 0.75.

Classification Based on Interaction With Slider

The research team then analyzed how participants engaged with the sliders during the entirety of the masking level task. Finger path trajectories showed participants with tinnitus gradually increased the masking level to the final value, whereas malingerers made larger, more erratic adjustments before abruptly stopping on their masking level.

Slider position over time was analyzed in three different ways to classify tinnitus patients versus malingerers:

• Deep neural network—AUC, 0.67; accuracy, 49%
• Random forest model—AUC, 0.77; accuracy, 65%
• Logistic regression model—AUC, 0.84; accuracy, 77%

Classification Based on Both Approaches

Finally, the team considered together the final values of the sliders and the participants' interaction with the sliders during all three tasks:

• VAS—AUC, 0.72; accuracy, 70%
• Masking level—AUC, 0.68; accuracy, 65%
• Matching—AUC, 0.70; accuracy, 66%
• Combined performance—AUC, 0.88; accuracy, 81%

The combined performance analysis also took into account demographic information and participants' scores on the Tinnitus Handicap Inventory.

Potential Applications

Automated analysis of standardized rapid tinnitus tests should be able to inform decisions about accepting or denying claims of tinnitus disability. The new system could also be useful for selecting participants for clinical trials of tinnitus biomarkers or treatments.

view original journal article Subscription may be required

Learn more about the Department of Otolaryngology–Head and Neck Surgery at Mass Eye and Ear

Refer a patient to Mass Eye and Ear/Mass General Brigham