"Less where possible, full compliance where required":our research into a shortshort version for the Tinnitus Handicap Inventory (THI) has been published in the International Journal of Audiology 

Publication Alert An adaptive short-form THI for online tinnitus triage

One Monday morning, a patient begins filling out the digital medical history form in preparation for her appointment with the ENT specialist. She has been hearing a soft, high-pitched ringing sound for months. It starts off smoothly, but at one point she thinks: "It’s quite a lot, actually. So many questions about the ringing in my ear."
That’s exactly what we wanted to address—without compromising on safety or clinical usability.

We are proud that our research on this topic has now been published in the International Journal of Audiology: an adaptive, short version of the Tinnitus Handicap Inventory (THI-12/8/5) that asks only as much as is necessary. No more.

The problem
Tinnitus is distressing, but the severity of symptoms varies from patient to patient. ENT specialists often measure the severity of tinnitus using the THI, a 25-item questionnaire with a total score of 0–100. If you do not differentiate according to symptom severity, everyone receives the same extensive THI questionnaire—often unnecessarily. Over-questioning reduces patients' willingness to fill out questions again during a follow-up measurement or subsequent intake.

The simple idea: make the THI adaptive
Using a genetic algorithm, we developed a three-step approach for the THI:
Step 1: 12 items. If the predicted THI total score ≤16 (mild), the questionnaire stops.
Step 2: 8 additional items. At 18–36 (mild), we add a short set and then stop.
Step 3: (remaining 5 additional items). If the score is above 36, the ConsultAssistant automatically asks all remaining items of the THI.

This is how THI-12/8/5 comes about: fewer questions where possible, complete transparency where necessary.

"Digital triage is designed to ease the burden on both patients and the healthcare team. It should be used for mild cases of tinnitus and become fully implemented as soon as the symptoms worsen."

— Niels Sprangers, Head of Analytics, ConsultAssistent

What our research shows
The publication highlights four points that are important in clinical practice:

  1. Low projection error: standard deviation ≈ 2.1 THI points; maximum deviation ≈ 8.5.
  2. High category agreement around the thresholds (light/mild/moderate/severe/catastrophic), sensitivity/specificity ≥90%.
  3. Profit depends on case mix:
    1. Dutch holdout: average 19.8/25 items (≈21% less).
    2. Light (step 1): 12/25 items (≈52% less).
    3. Tertiary population (heavier patients): less reduction — safety remains assured because more patients complete all 25 items.
  4. Algorithmic shortening works more broadly: the principle (genetic algorithm-driven selection) is effective and simple and can potentially also be applied to other questionnaires.

Important: we use and report the total THI score for triage; we do not interpret subscales. The adaptive design does not replace the THI-25: a projection >36 always results in the full questionnaire being administered.

How this looks in practice
– THI ≤16 (mild) → 12 items; explanation/reassurance, routine audiological care, self-management.
– THI 18–36 (mild) → 20 items; step-by-step care (e.g., hearing interventions, brief CBT-oriented self-help) and reassessment after 6–12 weeks (think MCID ≈7 points).
– THI >36 → automatic THI-25; consider multidisciplinary audiological and/or psychological support.

Why shortening is important now
Digital care pathways are part of everyday practice. Patients prepare at home; teams want to use that information in a targeted manner. An adaptive questionnaire that adjusts to the severity of symptoms fits in seamlessly with this. Asking too many questions reduces response rates and adoption; targeted shortening increases usability.

Standard in our digital ENT medical history
THI-12/8/5 is built into ConsultAssistent and integrated with various EHRs (including HiX, Nexus, and Emma). Implementation is primarily a matter of configuration: fixed thresholds, clear reporting, and also an AI summary as a pre-consultation report, based on the entire digital medical history.

As far as we are concerned: the new standard
Not every patient needs to answer every question to receive good care. And no patient should have to fill in too little information when there is more going on. With THI-12/8/5, we do this in our digital medical history for ENT — and we use data to show that it is possible.

Want to read more or find out more?
– See here the article in the International Journal of Audiology
– Are you interested in the full article? Contact us for an e-print of the full article. Send an email to research@consultassistent.nl
– Would you like to see how this works in the digital ENT medical history and how it fits into your care pathway? Contact us for a demo

Performance depends on the case mix; in tertiary populations, the reduction is smaller. Above 36, the THI-25 is administered in full. The short form is intended for triage and reports the total score.