METHOD
We sought to reproduce background noise in the clinical context by adjusting the signal to noise ratio (SNR) during testing. We chose adaptive SNR, using the Bamford-Kowal-Bench (BKB) Sentence Lists read by a clinician14 whilst a PARROT machine15 produced the background babble noise (simulated speech such as you might hear in a crowded pub or emergency department) at predetermined levels of noise. The PARROT is a portable digital speech screening system used to assess speech discrimination using a range of recognised speech discrimination tests.
The BKB sentences used in this test were published in 1979 as a protocol for testing hearing impaired children and developed as a SiN test by Niquette et al, 200316. There are 10 sentences in each list with 18 lists. Each sentence has three or four words which must be repeated by the subject.
A percentage score can be given for how many words are correctly repeated.
To determine background noise levels in our hospital we conducted 2x 30 second sound meter recordings (ATP Digital sound level meter/ 8928, [calibrated by NHS audiometric calibration service, Audiology department, Withington community Hopsital, Manchester]) in 4 discrete environments; The office, the emergency department (ED), the intensive care unit (ITU) and the operating theatre during normal, daylight working hours.
Background noise level minimum and maximums (during daylight hours with regular levels of staff) were recorded as follows:
40-55dB Office 48-66dB ED 50-78dB ITU 53-84dB Theatre
Five candidates representing our hospital ENT department were chosen, 2 women and 3 men. Age range was 29-49 years. Median age of candidates was 39 years old.
Initial 250Hz – 8kHz Pure tone audiograms were conducted to confirm no significant hearing loss in our 5 candidates, who had no previous otology history or significant co-morbidity.
All testing was conducted in a soundproofed audiometry booth. Baseline Standard BKB sentences were conducted in silence for all candidates without PPE. Scores were 100% for all candidates.
We then conducted BKB sentence testing whilst each subject wore the facial PPE suitable for aerosol generating procedures (AGPs) (fit-tested FFP3 mask and head visor). All subjects had previously undergone fit testing to ensure that the PPE worn was fitting appropriately for each individual.
The researcher spoke using the BKB word lists whilst wearing AGP PPE. The subject wore the same PPE at a distance of 2m. The PARROT machine was placed behind and above the head of the researcher (Parrotplus 2, produced by Soundbyte). Background noise (adult) babble settings were chosen as characteristic values representing different environments as follows:
45 dB Office 55 dB ED 65 dB ITU 70 dB Theatre
Each candidate underwent BKB testing at the 4 background noise levels, in three test conditions:
a) Candidate and researcher in normal conditions without PPE, normal researcher voice b) Candidate and researcher both in AGP PPE, normal researcher voice c) Candidate and researcher both in AGP PPE with researcher attempting to raise voice.
Raised voice was deemed researcher voice raised to the point at which the researcher felt their voice was comprehensible against background noise.
The percentage of key words in the BKB sentences repeated by the candidate was recorded. Each sentence was read once by the researcher and was not repeated.
During day-to-day working and conversation, people do not speak at the same intensity during the whole conversation and background noise also fluctuates rather than remaining at constant levels, hence our decision to use live, fluctuating voice rather than pre-recorded voices amplified to a fixed and constant volume17. When speakers adjust their voice to overcome background noise, this is known as the Lombard effect. Although attempting to raise one’s voice or shout usually causes only a small increase in volume between 5-10dB18, we chose to measure the volume of voice produced by the researcher as a secondary outcome measure. This was not our primary concern as attempts to raise one’s voice in day to day clinical practice will have both inter and intra-person variability, so we felt the integrity of the simulation was preserved regardless of actual volume levels produced by the researcher.