Journal Logo

Research Article

Improving the Differential Diagnosis of Otitis Media With Effusion Using Wideband Acoustic Immittance

Merchant, Gabrielle R.1; Al-Salim, Sarah2; Tempero, Richard M.3; Fitzpatrick, Denis1; Neely, Stephen T.1

Author Information
doi: 10.1097/AUD.0000000000001037
  • Free
  • Editor's Choice

Abstract

INTRODUCTION

Otitis media (OM) is among the most common childhood illnesses. There is wide variability in the characteristics and sequelae of a given episode of OM. Effusions present in OM may be serous, mucoid, or purulent, and vary considerably in both viscosity (Takeuchi et al. 1989) and volume (Koivunen et al. 2000). OM with effusion (OME) is characterized by the presence of a non-infected effusion in the middle ear cavity, with at least 25% of episodes of OME persisting for ≥3 months (Rosenfeld & Kay 2003). OME is often accompanied by a fluctuating and transient conductive hearing loss (CHL), but the degree of this CHL can vary from 0 to 40 dB HL (Gravel et al. 2006).

One significant concern regarding OME is that it is often asymptomatic, resulting in a delay in diagnosis and management. There are substantial periods of time where OME can go undetected, and these periods of time often coincide with critical stages of growth and development, including speech and language development (Hartley & Moore 2005; Di Francesco et al. 2016). Consequently, there is a significant potential negative impact of OME, particularly for OME with associated hearing loss, on the development of speech, language, and cognition in children (Rosenfeld et al. 2016). OME with associated hearing loss may also cause longer-term deficits in aspects of auditory processing, like binaural hearing, due to fluctuating and asymmetric auditory deprivation during these critical developmental periods (Whitton & Polley 2011).

Current assessment of OME includes otoscopy, single-frequency tympanometry, and behavioral assessment of hearing. These measures can often (but not always) determine the presence or absence of OME or middle ear dysfunction but provide limited further differentiation. Otoscopy (both standard and pneumatic) relies on the subjective interpretation of visual inspection of the external ear and tympanic membrane (TM) by the physician. Diagnosis and interpretation accuracy of otoscopy is highly variable, largely dependent on physician expertise, and has been shown to range from 40 to 70% (Blomgren & Pitkäranta 2003; Pichichero 2003; Pichichero & Poole 2005). In addition, gaining a clear unobstructed view of the TM in pediatric patients is not always possible, and given there is no active infection present, signs of OME are not always readily obvious (Sassen et al. 1994). Single-frequency tympanometry is often used to identify OME in a more objective manner, as it is simple and easy to measure. The sensitivity and specificity of tympanometry in diagnosing OME has been shown to be 86% and 72%, respectively, with an overall accuracy of 84% (Anwar et al. 2016). However, single-frequency tympanometry is limited in detecting minor changes in middle ear mechanics, and the presence of abnormal single-frequency tympanometry does not necessarily indicate the presence of a CHL as it has poor test performance for predicting the degree of CHL (Margolis et al. 1994; Vlachou et al. 2001; Baldwin 2006). Behavioral audiometry is utilized to determine hearing thresholds, and thus degree of CHL, in children with OME. However, behavioral audiometric techniques require the attention and compliance of children to respond repeatedly to threshold-level stimuli. Obtaining ear-specific air and bone conduction thresholds to accurately quantify hearing can be challenging in young children where OME is most prevalent, often resulting in limited, incomplete, and/or unreliable information regarding hearing status (Hunter et al. 1994).

If a full behavioral audiologic evaluation cannot be obtained in a child with OME, whether and how hearing is affected by a middle ear effusion cannot be determined. Determination of the degree to which hearing is affected in a given episode of OME is important as it should be considered in management decisions, such as whether to monitor the middle ear status over time or to proceed with surgical insertion of tympanostomy tubes. Knowledge of the amount of hearing loss present can help predict whether an episode of OME will spontaneously resolve, as children with OME with a pure-tone average of ≥30 dB HL in the better ear tend to have episodes of OME that persist rather than resolve (MRC Multi-Centre Otitis Media Study Group 2001). Accordingly, many professional organizations have clinical practice guidelines that include hearing loss criteria, typically hearing loss of 25 to 40 dB HL or greater, for surgical treatment (Chow et al. 2019). Current clinical best practice guidelines codeveloped by the American Academy of Otolaryngology—Head and Neck Surgery Foundation, the American Academy of Pediatrics, and the American Academy of Family Physicians also state that hearing loss should be taken into account when considering the diagnosis and management of OME (Rosenfeld et al. 2016). Our lack of an alternative method to behavioral audiometry to determine or estimate the CHL caused by a given episode of OME limits evidence-based management decisions as to whether, how, and when to treat OME.

Recent work from our laboratory demonstrated that the volume of an effusion present in the middle ear space directly impacts how much hearing loss occurs as a result of a given episode of OME (Al-Salim et al., 2021). Children that had ears with full effusions, or middle ear effusions that filled the entire middle ear cavity, generally had a flat moderate CHL with a mean 4 frequency pure-tone average of 43 dB HL in those ears. In contrast, ears with partial effusions or ears that were clear (no effusion) at the time of tube placement generally had a slight or mild CHL, with mean pure-tone averages of 22 and 17 dB HL, respectively. Age-matched normal hearing children demonstrated hearing thresholds within normal limits, with a mean pure-tone average of 7 dB HL. These results are consistent with temporal bone studies that have shown that even small amounts of air present in a middle ear cavity that is otherwise filled with fluid can greatly improve sound transmission, resulting in substantially less hearing loss, and it is not until the middle ear cavity is completely full of an effusion that considerable attenuation of the signal occurs (Ravicz et al. 2004). Therefore, while an effusion may be present and result in abnormal pneumatic otoscopy or tympanometry, the auditory signal transmitted to the brain may be unaffected or only minimally affected if the middle ear cavity is not full. This may also explain why it has been reported that as few as 15% of children with OM demonstrate hearing thresholds greater than 25 dB HL (Gravel & Wallace 2000).

Given this relationship between effusion volume and hearing loss, an objective method to predict average thresholds in children with OME would be useful for instances when thresholds cannot be obtained behaviorally. Physiologic tests in the audiology clinic are commonly utilized to obtain objective information regarding the functioning of various parts of the peripheral auditory system, especially in young children, where behavioral data may be limited, unreliable, or unobtainable. However, as described above, the commonly used physiologic test to detect OME (tympanometry) only has an overall accuracy of 84% and is not sensitive to small changes in middle ear mechanics. This is consistent with what was observed by Al-Salim et al. 2021, as there was no systematic effect of effusion volume on outcomes for standard 226 Hz tympanometry. Thus, there exists a need for a diagnostic measure that can provide information on effusion characteristics, specifically effusion volume.

Wideband acoustic immittance (WAI) is a promising tool that has been studied considerably in the research literature, but which has not yet been adopted into clinical practice, that may fulfill the aforementioned requirements. WAI refers to a group of measurements (i.e., absorbance, power reflectance [PR], admittance, and impedance, among others) used to represent the acoustic behavior of the ear. These measurements are made in the ear canal in response to wideband stimuli (e.g., a click or a chirp from 200 to 6000 Hz or 8000 Hz) and compare a sound input to the absorbed or reflected portions of that sound. WAI can be measured at ambient pressure or at varying positive and negative static pressures (often referred to as wideband tympanometry). Measurement of WAI at varying static pressures allows for assessment of WAI at tympanometric peak pressure (TPP). Previous work has demonstrated the influence of static pressure on WAI (e.g., Voss et al. 2012), and measuring WAI at TPP theoretically removes the influence of any middle ear static pressure occurring in a given patient on a given day on WAI. This can help remove the effect of static pressure on the WAI responses and isolate other nonstatic pressure-related mechanical changes in the system (Keefe & Levi 1996; Margolis et al. 1999; Keefe & Simmons 2003; Liu et al. 2008; Keefe et al. 2012; Sanford et al. 2013; Sun 2016).

As the impedance of the system changes due to pathologic alterations in middle- and inner-ear mechanics, changes in WAI can, in many cases, be demonstrated. WAI has been shown to have clinical value in differentiating origins of CHL, including both middle- and inner-ear causes (Feeney et al. 2003; Allen et al. 2005; Feeney et al. 2009; Shahnaz et al. 2009a,2009b; Nakajima et al. 2012; Voss et al. 2012; Nakajima et al. 2013; Prieve et al. 2013; Merchant et al. 2015; Merchant et al. 2016,Merchant et al. 2019). In addition, WAI is sensitive to the presence of middle ear effusion in infants and children with OM (or CHL due to suspected OM) (Keefe & Simmons 2003; Hunter et al. 2008; Beers et al. 2010; Keefe et al. 2012; Ellison et al. 2012,Prieve et al. 2013; Hunter et al. 2013; Sanford & Brockett 2014; Voss et al. 2016; Won et al. 2020), in adults with OM (Feeney et al. 2003), in temporal bone simulations of middle ear effusion (Voss et al. 2012), and in animal models of acute OM (Guan et al. 2017). However, though WAI responses in ears with middle ear effusions often differ from normal, there is also wide variability, hypothesized to possibly be due to the type and/or amount of effusion. Temporal bone simulations have, indeed, demonstrated that at least one measure of WAI, PR, is sensitive to not only the presence of middle ear effusion, but also the volume of effusion, suggesting that WAI may provide enough sensitivity to detect variability in OM effusion characteristics (Voss et al. 2012). In addition, Won et al (2020) found evidence of possible systematic variability in WAI in a small set of ears in children with middle ear effusion thought to vary in amount and viscosity. In this study, characteristics of middle ear effusions were determined by optical coherence tomography (OCT) (i.e., scant versus severe for volume and/or serous versus mucoid for viscosity), and trends were observed in WAI depending on effusion characteristics. One limitation of this work is that there was no surgical confirmation of effusion characteristics, and OCT classifications were in contrast with physician diagnoses in several instances. Though this may indicate that OCT is more sensitive than a physician diagnosis and the lack of surgical confirmation allows accessibility to less severe cases of OME, the sensitivity and specificity of OCT to define these categories is still under development, and additional work in this area is needed.

The goal of this work is to determine whether there is a systematic effect of effusion volume on WAI. A number of responses can be calculated from WAI measurements. These include, but are not limited to, impedance, pressure reflectance, PR, and absorbance. Pressure reflectance R(f) of the ear is defined as the ratio of reflected pressure to forward pressure. Pressure reflectance is related to the impedance at the point of measurement by the following equation:

R(f)=(Z(f)Z0)(Z(f)+Z0)s

where f is frequency, Z(f) is the frequency-dependent impedance looking into the ear canal, and Z0 is the characteristic acoustic impedance of the ear canal at the measurement point. PR is the squared magnitude of pressure reflectance |R( f )|2 and is a number between 0 and 1, where PR = 0 suggests that all power is transmitted to the middle ear and where PR = 1 represents that all power is reflected back into the ear canal at the TM. Absorbance is the inverse of PR and represents the sound power absorbed by the system near the TM (Absorbance = 1 − PR). Absorbance and PR are appealing measures of WAI to study as they are relatively insensitive to location of the probe tip in the ear canal (Stinson 1990; Huang et al. 2000; Voss et al. 2008), increasing clinical utility and feasibility. Absorbance also shares similar properties to admittance with respect to interpretation of the response and has thus become the most common outcome parameter reported in translational studies of WAI. Therefore, the main outcome parameter utilized in this work is absorbance. We hypothesize that absorbance will decrease as the volume of effusion present in an ear increases. If there is a systematic effect of effusion volume on WAI, WAI could be utilized to predict thresholds in children with OME based on their effusion volume.

MATERIALS AND METHODS

Study Population

Children With OME

The OME group consisted of 26 children (14 males and 12 females) recruited from the clinics of 2 otolaryngologists at Boys Town National Research Hospital. All children had a history of chronic or reoccurring OM and were scheduled for tympanostomy tube placement in at least 1 ear. All children had a current diagnosis of OME and at least one ear with a middle ear effusion, as confirmed by their otolaryngologist, upon enrollment in the study. Of the 26 children who participated, 23 were scheduled for bilateral tube placement and 3 for unilateral tube placement, resulting in a total of 49 ears. Participants ranged in age from 9 months to 11 years, 2 months with a mean age of 35.65 months (SD = 28.98, median age = 23 months, interquartile range = 33.25 months). Participant race was reported as white for 21 participants, black for 3 participants, more than one race for 1 participant, and with the response declined for 1 participant. Participant ethnicity was reported as Non-Hispanic or Latino for 22 participants, Hispanic or Latino for 3 participants, and declined for 1 participant. Sensorineural hearing loss was ruled out either through normal audiologic testing results before tube placement or at postoperative testing sessions. Children were from families where English was spoken fluently by at least one primary caregiver. Participants had no history of permanent hearing loss or significant developmental delays. There was no history of previous tube placement in all ears with the exception of four ears from three children.

Normal Hearing Children With Healthy Ears

The control group consisted of nine children (four males and five females) with no recent history (within the past 12 months) of OM. Children were recruited from the Boys Town National Research Hospital participant recruitment database and ranged in age from 10 months to 10 years, 11 months with a mean age of 44.11 months (SD = 38.85, median age = 33 months, interquartile range = 37 months). Participant race was reported as white for eight participants and more than one race for one participant. Participant ethnicity was reported as Non-Hispanic or Latino for eight participants and Hispanic or Latino for one participant. These children had no history of ear surgery, including placement of tympanostomy tubes. Normal control group participants were required to have type A tympanograms and audiometric thresholds of 20 dB HL or better from 500 to 8000 Hz in each ear included for participation. If ear-specific audiometric thresholds could not be obtained, then present DPOAEs were required from 1500 to 8000 Hz in each ear included in analyses. Fourteen ears met the criteria for inclusion in the control group (three ears were not included due to significant negative TPP and one ear was not included as the participant only tolerated testing in one ear).

Informed consent was obtained from all participants and caregivers for testing procedures approved by the Institutional Review Board at Boys Town National Research Hospital.

Wideband Acoustic Immittance

WAI was measured in both ears for each participant using a Titan IOWA probe (Interacoustics, Middelfart, Denmark). Children were seated upright in their caregiver’s lap or independently in a chair for testing. Children were distracted during testing by watching silent videos on an iPad or by the tester or a test assistant displaying and playing with various toys. For children with OME, measurements were completed in the child’s pre-operative patient room (by author G.R.M.) the morning of surgery immediately before tympanostomy tube placement. For OME participants undergoing bilateral tympanostomy tube placement, the first ear measured was chosen at convenience based on how the child was positioned and comfortable when testing began. For normal control participants, measurements were completed in our laboratory, with left ear measurements generally being completed before right ear measurements, though participant compliance sometimes resulted in right ear measurements being completed first.

Custom software (MATLAB; MathWorks, Inc., Natick, MA) written by author D.F. and partly based on code shared as part of the Titan Research Platform (Interacoustics) was utilized for both WAI daily calibration and data collection. To obtain the source pressure and impedance of the system, a calibration procedure (e.g., a Thévenin calibration) was completed daily before each test session in a four-cavity calibrator (Interacoustics) according to the Titan Research Platform specifications and published methods (Nørgaard et al. 2017).

WAI was measured in response to a wideband click stimulus (220 to 8000 Hz) in both an ambient and pressurized tympanometric condition in each ear, the latter of which will be referred to as wideband tympanometry. All participants underwent pressurized wideband tympanometry testing, while only a subset of participants underwent ambient-pressure testing. This was due to both patient compliance and the fact that ambient testing was not yet available in our custom software when data collection began. If both ambient and wideband tympanometric testing were completed on the same ear, ambient measurements were always completed first. On many occasions, the probe was removed and reinserted between these two measures to allow participants a break in testing (or because the participant pulled out the probe). Wideband tympanometry measures were completed with downswept pressure ranging from +200 to −300 daPa at a rate of approximately 100 daPa/s. The absorbance response at (or nearest to) 0 daPa (pressurized “ambient”) and TPP were extracted from the wideband tympanogram and utilized in this work. TPP was calculated as the pressure of the maximum low-frequency absorbance passband from 215 to 2000 Hz. This is the default method for calculating wideband TPP for the Interacoustics Titan device and is generally consistent with published methods (e.g., Liu et al. 2008). We report absorbance at ambient pressure from the ambient recording as well as absorbance at ~0 daPa (pressurized “ambient”) and TPP from the tympanometric response here.

Responses were inspected in both real-time and post-processing for the presence of air-leaks (using criteria suggested by Groon et al. 2015). Measurements were repeated if indications of leaks were present and noted at the time of the measurement and as needed due to participant noise (e.g., if the patient yawned or spoke during the recording or if a proper seal was not achieved). A minimum of two qualitatively identical responses were collected for each ear to ensure reliability of the response, and a single tracing of these two was chosen at random for analysis. Responses were smoothed across frequency in 1/6th octave bands.

Tympanometry

Traditional 226 Hz tympanometry was also measured in each ear with the same device using the Interacoustics Titan Suite Clinical Software platform and the IMP440 module, for the purpose of comparison. This was completed at the same time as the WAI measures in the patient’s preoperative room. The probe was removed and reinserted if tympanometry and WAI measurements were made back to back, though it was common to take a short break between measures. Tympanograms were measured using a 226 Hz probe tone (as all participants were older than 9 months of age) and a positive to negative pressure sweep from +200 to −300 daPa, with volume, admittance, pressure, and gradient automatically calculated by the manufacturer software. Tympanograms were classified into types A, B, and C (Jerger 1970) based on the criteria used by Alaerts et al. (2007). Type A tympanograms were peaked with a peak pressure ≥−150 daPa and an admittance >0.2 mL. Type B tympanograms lacked a distinct peak, had admittance <0.2 mL, or a tympanometric width >200 daPa. Type C tympanograms had a peak pressure <−150 daPa and an admittance >0.2 mL.

Middle Ear Effusion Volume Assessment

For ears with OME, the volume of the middle ear effusion was determined during surgical insertion of tympanostomy tubes in the operating room. Subjective assessment of effusion volume was made via binocular surgical otomicroscopy immediately before myringotomy (when the TM was intact) with confirmation after tube placement by the otolaryngologist performing the procedure. (Note that the author R.M.T. was the otolaryngologist for 46 of the ears included in this work, with 3 ears from 2 participants included from a second collaborating otolaryngologist. Both otolaryngologists used the same procedure for classification and discussed their technique to ensure classifications were made as consistently as possible.). A subjective estimate of the volume of the effusion was made by describing the amount of effusion in relation to the middle ear cavity size as clear, ¼ full, ½ full, ¾ full, or full. The ¼, ½, and ¾ full judgments were all combined into a “partial” category for analysis. The otolaryngologist was blinded to preoperative testing results so as not to bias their assessment.

Attempts were also made to objectively quantify the volume of the effusion; however, the reliability of these assessments was poor and therefore deemed unsuitable for analysis. The objective volume assessments were unsuitable for two reasons: (1) the entire effusion could not be suctioned from the middle ear space and some of the effusion remained in the suction tubing and the tubing of the Medtronic Juhn Tym-Tap Middle Ear Fluid Aspirator/Collector (Medtronic, Minneapolis, MN) device, thus significantly limiting the ability to quantify the amount suctioned accurately and (2) the absolute volume may not actually be the value of interest, given that middle ear cavity size can vary and the impact of the effusion on the auditory system may be affected by the percentage of the cavity filled and not the absolute volume of the effusion (Ravicz et al. 2004; Voss et al. 2012). Thus, even precise quantification of the absolute volume of the effusion sample may not be as informative as the subjective assessment of how full the middle ear cavity was.

Analyses

To evaluate the clinical utility of absorbance measures in the differential diagnosis of OME, a multivariate logistic regression approach was utilized. Given each absorbance response contains data at 364 frequency points, we first reduced the dimensionality of the data by using a principal component analysis (PCA) on the entire dataset. This PCA approach reduces the likelihood of overfitting the data and removes the issue of collinearity because the principal components are orthogonal. We retained the three most significant principal components, as this resulted in the best performance as determined by the largest area under the receiver operating characteristic curves (AUCs, as detailed further below).

After the data were reduced, a multivariate logistic regression (mnrfit, MATLAB) based on the three principal components was completed. We used a binary two-category approach, first classifying absorbance responses as effusion present (full and partial ears) or effusion absent (clear and normal control ears), and then classifying effusion-filled ears as either full or partial, and ears without effusion as either clear or normal controls. For the regressions completed on the subsets of the data (i.e., full versus partial and clear versus normal control), a PCA was completed on that subset of the data prior to the regression, again retaining the three most significant principal components. The output of this regression analysis was a prediction as to which OME category an individual absorbance response belonged to. Regression test performance was assessed by calculating the sensitivity, specificity, accuracy, and AUC for the three binary classifications, with positive detection being the effusion present when compared to effusion absent, detection of full ears when compared to partial, and detection of clear ears when compared to normal control ears. AUC was calculated as the average sensitivity (or hit rate) as a function of 1-specificity (or false-alarm rate). An AUC of 0.7 to 0.8 is generally considered acceptable, 0.8 to 0.9 considered excellent, and more than 0.9 considered outstanding (Mandrekar 2010).

Finally, to simulate how this approach would perform on novel unseen data, as would be more realistic of what we might expect if it were to be utilized in clinic, we sought to evaluate how the regression would perform if the model was trained on a subset of the data (70%) and then validated with the remaining unseen data (30%). To accomplish this, the dataset was randomly separated into a training set and a validation test set. The training set was constrained to ensure that 70% of each effusion group was included in the training set, which also ensured that each group was evenly represented in the validation set. The regression model was trained on the training set (as described above, with the PCA and regression being completed for the 70% training data) and then evaluated for the 30% of unseen data for each of the 3 binomial regressions (effusion present versus absent, partial versus full, and clear versus normal control). The split between training and validation sets was randomized and repeated 1000 times to improve the reliability of the reported average AUCs.

We also compared absorbance data collected at true ambient pressure to absorbance data extracted from the wideband tympanogram at ~0 daPa. To determine whether absorbance at ambient and 0 daPa data were statistically different, absorbance data were averaged into 1/3rd octave bands from 220 to 8000 Hz, reducing the number of data points from 364 to 16 (such as in Prieve et al. 2013), and paired-samples T-tests were completed.

Principal component and multivariate linear regression analyses were completed using MATLAB R2019a (MathWorks, Inc.). Paired-samples T-tests were completed using SPSS Statistics v. 25 (IBM Corp., Armonk, NY). Figures were created using IGOR Pro 8 (WaveMetrics, Inc., Portland, OR).

RESULTS

Middle Ear Effusion Volume Assessment

Of the 49 ears with a diagnosis of OME included in our study, 18 (37%) were confirmed to have full effusions, 13 (26%) were confirmed to have partial effusions, and 18 (37%) were confirmed to be clear of effusion at the time of tympanostomy tube placement. Note that the average number of days between diagnosis of OME in clinic and the assessment of WAI and surgery where the volume assessment took place was 18 days (range 6 to 43 days). Of the four ears with a previous history of tube placement, two ears, one each from two different participants, had partial effusions and two ears, both from the same participant, had full effusions.

Although not of primary interest in this study, effusion viscosity was subjectively characterized by the operating otolaryngologist as mucoid, serous, or purulent. All 18 ears with full effusions and 11 ears with partial effusions had nonpurulent mucoid effusions, with only 2 ears (both partial effusions) demonstrating nonpurulent serous effusions, and 0 ears demonstrating purulent effusions. Given this lack of variability, effusion viscosity was not analyzed or further considered in this work.

Standard Tympanometry

All but two ears with effusions present at the time of tympanostomy tube placement (i.e., full and partial ears) demonstrated type B tympanograms. One ear with a full effusion had a type C tympanogram and one ear with a partial effusion had a type A tympanogram. Of the 18 ears that were clear, 6 had type B tympanograms, while 11 had type A and 1 had a type C tympanogram (Table 1). All normal control ears were required to have type A tympanograms to be included in the study. Thus, consistent with previous findings (Al-Salim et al., 2021), there was no consistent effect of effusion volume on standard 226 Hz tympanometry.

TABLE 1. - Number of ears for each type of tympanogram based on the criteria of Alaerts et al. (2007) for full, partial, and clear ears
Tympanogram Types
A B C
Full ears 0 17 1
Partial ears 1 12 0
Clear ears 11 6 1

Absorbance

Absorbance at TPP and at 0 daPa were extracted from the pressurized wideband tympanometry responses and grouped by surgically confirmed effusion volume category. Figure 1 displays the mean (±the standard error of the mean, SE) for each of the 3 effusion groups (full, partial, and clear) as well as for the normal control group at TPP and 0 daPa. Data from the normal control ears are consistent with published data in this age range (e.g., Hunter et al. 2008; Beers et al. 2010; Ellison et al. 2012; Keefe et al. 2012). Absorbance patterns between groups at TPP and 0 daPa are similar. Results demonstrate that as categorical effusion volume increases, absorbance generally decreases, with obvious qualitative separation between the 3 effusion groups from 500 to 6000 Hz. Clear ears and normal control ears demonstrate overlap at many frequencies. However, clear ears have reduced absorbance as compared to normal control ears from 750 to 2000 Hz.

Fig. 1.
Fig. 1.:
Mean (±standard error of the mean) absorbance for 18 ears with full effusions (solid), 13 ears with partial effusions (dashed), 18 ears clear of effusion (dotted), and 14 normal control ears (dot-dashed) at tympanometric peak pressure (TPP; left, red) and 0 daPa (right, black).

A subset of our participants in all four groups (full, partial, clear, and normal control) completed absorbance testing at both ambient pressure and under pressurized conditions. We explored how wideband absorbance at ambient pressure and 0 daPa varied in all 4 of our participant groups. Figure 2 (left) shows mean absorbance data for ears in each of our 4 groups at 0 daPa (pressurized “ambient”) extracted from the wideband tympanogram and at true ambient pressure. Figure 2 (right) also displays the mean frequency-by-frequency differences between individual responses at ambient and 0 daPa in the same ear for all ears shown in the left panel (ambient minus 0 daPa). Qualitatively, the mean responses by group are largely overlapping. There is a visible trend for ambient absorbance to be slightly higher than 0 daPa absorbance in full ears around 4 kHz. Ambient and 0 daPa absorbance data were averaged into 1/3rd octave and paired-samples T-tests indicated no significant differences between the ambient and 0 daPa absorbance in any 1/3rd octave band for any of the 4 groups.

Fig. 2.
Fig. 2.:
Left, Mean (±standard error of the mean) absorbance at 0 daPa (black) and ambient (teal) for 9 ears with full effusions (solid), 8 ears with partial effusions (dashed), 12 ears clear of effusion (dotted), and 14 normal control ears (dot-dashed). Right, Mean frequency-by-frequency differences between individual responses at ambient and 0 daPa in the same ear for all ears shown in the left panel (ambient minus 0 daPa).

Individual absorbance data at both 0 daPa and TPP for each effusion volume group are displayed in Figure 3. Responses at TPP and 0 daPa are similar for full ears. In the 3 groups with partial or complete aeration of the middle ear cavity (i.e., partial, clear, or normal control ears), there is a trend for absorbance at TPP to be higher than absorbance at 0 daPa.

Fig. 3.
Fig. 3.:
Mean (±standard error of the mean) absorbance at tympanometric peak pressure (TPP, red) and 0 daPa (pressurized “ambient”) with individual data overlaid for 18 ears with full effusions (top left), 13 ears with partial effusions (top right), 18 ears clear of effusion (bottom left), and 14 normal control ears (bottom right).

Given the qualitative separation between the means of the groups, statistically significant differences in multiple frequency bands are highly likely. However, of particular interest to us was not these obvious group-level differences, but whether clinically significant and meaningful separation was present for individual measurements. To answer this, we used a multivariate logistic regression model to predict which category each individual absorbance response belonged to, and determined test performance by calculating the sensitivity, specificity, accuracy, and AUC. We completed this analysis on the responses at TPP in an effort to isolate the effects of the middle ear effusions, as negative static pressure is commonly found in children with OME.

The first step of this analysis was to complete a PCA on the entire dataset. We retained the three most significant principal components from the PCA, as this resulted in the best performance as determined by the largest AUCs. The eigenvalues of these 3 components were principal component 1: 11.89, principal component 2: 1.28, and principal component 3: 0.73. The frequency weighting of the principal components (also known as the loadings or the principal component coefficients, which are demonstrated as a function of frequency) is shown in Figure 4. These weightings help identify which frequency points loaded on to each principal component as well as the direction of that loading, providing insight into which frequency regions within the absorbance responses contributed the most to each of these three principal components.

Fig. 4.
Fig. 4.:
Principal component weight by frequency for the 3 principal components (PC1, PC2, and PC3) included in the multivariate logistic regression model for the entire dataset of 63 ears.

We then completed multivariate logistic regressions based on these three principal components. We used a binary two-category approach, first classifying absorbance responses as effusion present (full and partial ears) or effusion absent (clear and normal control ears), and then classifying effusion-filled ears as either full or partial, and ears without effusion as either clear or normal controls. The results of these regressions are shown in Table 2. The regression predicted whether effusion was present or absent with 100% accuracy. It also classified 100% of the ears with confirmed effusion correctly as being either full of effusion or partially full of effusion and classified 75% of the ears without effusion as ears that were clear at the time of tube placement (but with a very recent diagnosis of OME and confirmed effusion present) or a normal control ear. Figure 5 displays the regression vector by frequency for each of these three regression models. Each regression vector is a linear combination of the principal component weights and provides a visual aid in interpreting the relative contribution of each frequency component towards distinguishing binomial classifications. The regression for effusions being overall present or absent, or present effusions being full or partial, both demonstrate regression vectors that relied heavily on information contained in the absorbance responses between 1 and 3 kHz. In contrast, the regression vector for classifying clear ears from normal control ears, when plotted on the same scale, appears nearly flat, suggesting that there was no strong association with a frequency region that could effectively predict the category of a given ear. This provides insight as to why accuracy for the classification of clear versus normal control ears is lower than that for the other two regressions, which is also consistent with the amount of overlap observed in these two groups as demonstrated in Figure 1.

TABLE 2. - Confusion matrices of the multivariate logistic regression on the 3 most significant principal components from the PCA analysis completed on all 63 ears to classify effusion as present or absent (left), the 31 ears with effusion present to classify as effusion full or partial (middle), and the 32 ears with effusion absent to classify as effusion clear or normal hearing control ears (right)
Effusion Present vs. Absent Partial vs. Full Effusion Clear vs. Normal Ears
Overall Confusion Matrix Overall Confusion Matrix Overall Confusion Matrix
Accuracy: 100% Accuracy: 100% Accuracy: 75%
Present Absent Full Partial Clear Normal
Present 31 0 Full 18 0 Clear 14 4
Absent 0 32 Partial 0 13 Normal 4 10
PCA, principal component analysis.

Fig. 5.
Fig. 5.:
Regression vector by frequency for each of these three regression models: normal control ears vs. clear ears (green, dotted), partial ears vs. full ears (pink, dashed), and effusion absent vs. present (blue, solid).

Next, we wanted to explore how the regression would perform on novel unseen data, as this would be more realistic of performance in a clinical setting. To do this, the model was trained on a subset of the data (70%) and then validated with the remaining unseen data (30%), with this process being completed iteratively 1000 times. Average accuracy, sensitivity, specificity, and the AUC of the 1000 trials were calculated for each training and validation set, based on the confusion matrix output, the results of which are shown in Table 3.

TABLE 3. - Confusion matrix of the multivariate logistic regression on the 3 most significant principal components from the PCA analysis completed on a training set containing 70% of the data and then tested on a validation set containing 30% of unseen data
Effusion Present vs. Absent Partial vs. Full Effusion Clear vs. Normal Ears
Training Confusion Matrix Training Confusion Matrix Training Confusion Matrix
Accuracy: 100%, Sensitivity: 100%, Specificity: 100%, AUC: 1.000 Accuracy: 99%, Sensitivity: 100%, Specificity: 99%, AUC: 1.000 Accuracy: 75%, Sensitivity: 78%, Specificity: 72%, AUC: 0.852
Present Absent Full Partial Clear Normal
Present 22,000 0 Full 12,889 111 Clear 10,287 2713
Absent 0 22,000 Partial 44 8956 Normal 2945 7055
Validation Confusion Matrix Validation Confusion Matrix Validation Confusion Matrix
Accuracy: 95%, Sensitivity: 95%, Specificity: 95%, AUC: 0.988 Accuracy: 89%, Sensitivity: 89%, Specificity: 88%, AUC: 0.944 Accuracy: 65%, Sensitivity: 67%, Specificity: 62%, AUC: 0.689
Present Absent Full Partial Clear Normal
Present 8459 541 Full 8459 541 Clear 3642 1358
Absent 439 9561 Partial 439 9561 Normal 1788 2212
This was done iteratively 1000 times with 1000 different random combinations of training and validation sets and done for 3 separate 2-category comparisons: effusion present vs. absent (left) on the entire dataset, full vs. partial (center) on the subset of data with present effusions, and clear ears vs. healthy normal control ears (right) on the subset of data without effusion present. Average results for the 1000 iterations are shown.
AUC, area under the receiver operating characteristic curve; PCA, principal component analysis.

For the classification of effusion present versus absent, training performance was identical to performance on the entire dataset (Table 2) with 100% accuracy, and validation performance was slightly reduced but still very high (accuracy: 95%, sensitivity: 95%, specificity: 95%, and AUC: 0.988). For the classification of present effusions being full effusions or partial effusions, training performance was similar to performance on the entire dataset (Table 2) with 99% accuracy; however, validation performance was reduced to an accuracy of 89% (accuracy: 89%, sensitivity: 89%, specificity: 88%, and AUC: 0.944). Finally, for the classification of absent effusions being clear ears or normal control ears, training performance was similar to performance on the entire dataset (Table 2) with 75% accuracy, with validation performance reduced to 65% (accuracy: 65%, sensitivity: 67%, specificity: 62%, and AUC: 0.689).

DISCUSSION

The purpose of this study was to determine whether WAI is sensitive to the volume of middle ear effusion in children with OME. Overall, our findings suggest that WAI, and more specifically, absorbance, is a strong and sensitive indicator of the volume of a middle ear effusion in children with OME, at least with respect to the categorical volume groups defined in this work. Absorbance is systematically reduced as the categorical volume of the effusion increases when measured at both TPP and 0 daPa (Fig. 1). This reduction is present at most frequencies but is greatest in the frequency range from 1 to 5 kHz. This suggests that the middle ear is less efficient at absorbing sound energy across most frequencies, but particularly the mid-frequencies, as the volume of an effusion increases in children with OME. These results are consistent with recent work from our laboratory demonstrating that the volume of the effusion present in the middle ear space (characterized categorically in the same way as done here) has a significant and systematic effect on the degree of hearing loss caused by a given episode of OME (Al-Salim et al. 2021). Children that had ears with full effusions, or middle ear effusions that filled the entire middle ear cavity, generally had a flat moderate CHL while ears with partial effusions or ears that were clear (no effusion) at the time of tube placement had thresholds that ranged from within normal limits to a slight or mild CHL. As expected, no systematic effect of effusion volume was found on standard tympanometry.

Absorbance Results

Absorbance from the normal control ears in our study is consistent with previously published normative data in similar age ranges (e.g., Hunter et al. 2008; Beers et al. 2010; Ellison et al. 2012; Keefe et al. 2012). To our knowledge, no other study has investigated the effect of effusion volume in children with OME in this way for comparison purposes, with the exception of a small set of ears grouped based on OCT results (Won et al. 2020). However, if we group all OME ears together (Fig. 6), our absorbance responses from ears with OME are consistent with published absorbance data in children with OME (Fig. 6; Keefe & Simmons 2003; Hunter et al. 2008; Beers et al. 2010; Keefe et al. 2012; Ellison et al. 2012,Prieve et al. 2013; Hunter et al. 2013).

Fig. 6.
Fig. 6.:
Mean (±standard error of the mean) absorbance for all 49 ears with otitis media with effusion (OME; solid) and 14 normal control ears (dot-dashed) at tympanometric peak pressure (TPP) for the purpose of comparison to previous literature.

In addition, the systematic reduction in absorbance as categorical effusion volume increases is consistent with previous evidence in human cadaveric temporal bone preparations (Voss et al. 2012), in investigations of the amount of TM movement as classified by pneumatic otoscopy (Ellison et al. 2012), and the comparisons of “scant” versus “severe” effusions from a limited dataset (Won et al. 2020). Voss et al (2012) investigated the impact of systematic increases in fluid injected into the middle ear cavity on PR (the inverse of absorbance). They found that PR increased (and therefore absorbance decreased) as the volume of fluid they injected increased until the middle ear space was completely full. Ellison et al (2012) classified ears with OME in their study based on the subjective stiffness of the TM as determined by pneumatic otoscopy. They had a five-point rating scale from one (normal) to five (no movement). Similar to what we observe here with increasing fluid volume, they found a similar reduction in absorbance with increasing stiffness, though the separation was not as pronounced. It is plausible that the stiffness categories in that work could be related to the volume of the effusion, with increasing stiffness resulting from increased effusion volume. Won et al (2020) also found a trend for reduced absorbance in their “severe” ears (i.e., ears with a larger amount of effusion) as compared to their “scant” ears, consistent with our findings.

Comparison of Absorbance at Tympanometric Peak Pressure and 0 daPa

Absorbance responses at TPP and 0 daPa were similar for full ears, which is unsurprising given these ears are unlikely to have a true peak static pressure if there is no aeration of the middle ear cavity (although TPP was still defined by whatever the software determined was the pressure at the maximum absorbance in the low-frequency band). In contrast, in the 3 groups with partial or complete aeration of the middle ear cavity (i.e., partial, clear, or normal control ears), there was a trend for absorbance at TPP to be higher than absorbance at 0 daPa. This may be due, at least in part, to the fact that static pressure causes a decrease in absorbance (e.g., Voss et al. 2012). Considering absorbance at TPP as opposed to 0 daPa or ambient pressure helps remove the effect of static pressure on the absorbance response (Keefe & Levi 1996; Margolis et al. 1999; Keefe & Simmons 2003; Liu et al. 2008; Keefe et al. 2012; Sanford et al. 2013; Sun 2016). Thus, absorbance at TPP would be expected to be higher if static pressure was impacting the 0 daPa response, consistent with our results.

Comparison of Ambient Absorbance and Absorbance at or Near 0 daPa

In a subset of our participants, we also compared absorbance at 0 daPa extracted from the wideband tympanogram with absorbance at true ambient pressure. Absorbance at 0 daPa extracted from the wideband tympanogram is often referred to as “ambient” absorbance. However, absorbance at 0 daPa is measured under a pressurized condition. In addition, absorbance measured at ambient pressure generally includes averaging of multiple click or chirp responses in rapid succession, while this averaging does not occur in responses extracted from the wideband tympanogram given the added variable of the pressure sweep. Therefore, 0 daPa “ambient” measures may not the same as absorbance measured under true ambient conditions. Sun (2016) found that wideband absorbance obtained from adult ears at ambient pressure was lower than absorbance obtained at 0 daPa, which they hypothesized may be due to tympanometric preconditioning given that the ear is pressurized to +200 and responses are collected at various positive static pressures before the response collected at 0 daPa (Burdiek & Sun 2014). Understanding differences between 0 daPa measures and measurements at ambient pressure is important, particularly given that much of the previous literature in the field on normal and pathologic ears was collected at ambient pressure (including WAI data on ears with OME), and comparisons may vary due to this and/or require specific normative data of the same pressurized condition. Despite the differences between the 2 measures and the differences observed in some of the previous literature, our results demonstrate no significant differences between absorbance measured at 0 daPa extracted from the wideband tympanogram and true ambient absorbance in ears with OME or in normal control ears. Given the potential added information gained by measuring in the pressurized condition (e.g., ear-canal volume, absorbance at TPP, and the wideband tympanogram), pressurized WAI responses would provide the most clinical utility in these populations.

Diagnostic Utility of Absorbance Using Multivariate Regression

In an effort to understand the clinical significance and implications of our group level absorbance findings on an individual level, a two-category multivariate logistic regression approach was utilized to classify individual absorbance responses at TPP based on their effusion volume group. This regression approach first considered the entire dataset and classified ears as effusion present (full and partial ears) or absent (clear ears and normal control ears) and did so with 100% accuracy. This is highly significant, given it is a substantial improvement to the diagnostic accuracy of tympanometry, which has been reported to be 84% in ears with OME (Anwar et al. 2016). The same two-category multivariate logistic regression approach was then used to classify individual ears with effusion present (all partial and full ears) as either full or partial. This was also performed with 100% accuracy. The regression approach was finally applied to the ears without effusion present to determine performance for classifying normal control ears from ears clear of effusion, resulting in an accuracy of 75%. While this approach had more difficulty in separating individual ears clear of effusion from normal control ears, this separation is likely less significant in terms of clinical utility. Understanding whether effusion is present or absent has longstanding clinical value, and the further differentiation of present effusions being full or partial could provide valuable insight for a provider as to the hearing status of the child given that effusion volume is a significant driver of the degree of hearing loss in children with OME (Al-Salim et al. 2021). However, there may not be as strong of a clinical need to differentiate ears that are clear of effusion, given case history alone may provide insight as to whether a recent history of OME was present. Finally, it should be noted that we did attempt a four-category multinomial logistic regression approach to classify the entire dataset into their four respective groups within one regression; however, the performance was improved using the above two-category multi-regression approach described above, and thus that approach was adopted.

The classification performance of this regression suggests high clinical utility of absorbance in determining effusion volume group in pediatric ears with OME. An automated algorithm could be developed from this approach, furthering clinical utility. However, one limitation of this approach is that, though the regression performed well on this dataset, this method does not quantify performance on novel unseen data. Given new data collected in a clinical setting would be “unseen” to the model, knowledge of how this approach would perform on unseen data would be of great use in considering the clinical impact. Thus, regression performance was also explored when the dataset was split into a training set (70% of the data) and a validation test set (30% of the data) to simulate the performance of this approach on unseen data, providing insight into the true clinical utility of this method. Overall, this approach demonstrated high sensitivity and specificity for classifying ears as effusion being present or absent and as present effusions being full or partial effusions with areas under the curve ranging from 1 to 0.944. Response classification for effusion being present or absent for the unseen data as part of the validation set demonstrated 95% sensitivity and 95% specificity, which is a notable improvement compared to the sensitivity and specificity of tympanometry for OME reported in the literature (86% and 72%, respectively, in Anwar et al. 2016).

Despite the lack of effusion present in both clear ears and normal control ears, this approach was also able to classify the responses from these ears, but with a more moderate sensitivity and specificity. The differences observed here between ears clear of effusion and normal control ears may suggest that even when the middle ear space appears to be fully aerated, an ear with recent OME appears to have lingering pathologic processes, which may include residual effusion in the middle ear air space (such as within the mastoid air cell cavities), scarring, inflammation, presence of a biofilm, and thickening of the TM, that affect the mechanical properties of the system (Maw & Bawden 1994; Daly et al. 2003). This may give insight as to why ears clear of effusion at the time of tube placement, but with a recent history of abnormal otoscopy and OME, demonstrate poorer thresholds than age-matched normal-hearing control ears (Al-Salim et al. 2021; Hunter et al. 1994).

Implications for Audiometric Behavioral Threshold Estimation

Our results help explain why some children with OME demonstrate moderate CHL while others demonstrate audiometric thresholds within normal limits (Al-Salim et al. 2021; Gravel et al. 2006). The degree of hearing loss associated with OME is an important factor in management considerations in children with OME. However, behavioral audiometric testing in the age range where OME is most prevalent is challenging and time-consuming, and obtaining a complete set of reliable air and bone conduction thresholds to define the hearing loss is often not possible (Hunter et al. 1994). Thus, providers cannot always rely on knowledge of the presence and degree of hearing loss in making management decisions relating to OME, such as whether to watch and wait or suggest placement of tympanostomy tubes. A quick, reliable, objective measure indicating the volume of the effusion in a given episode of OME would have significant clinical utility, particularly in instances where behavioral audiometric data is unattainable. These findings suggest that WAI can do just that.

Limitations

A limitation of this study is that only one characteristic of the middle ear effusion, effusion volume, was able to be explored. The impact of viscosity or purulence could not be evaluated as nearly all ears with effusions included in this study were found to have non-purulent mucoid effusions. While this could be a sampling issue, it is also possible (if not likely) that nonpurulent mucoid effusions are more common in ears that are at the point of having clinical indications for surgical placement of tympanostomy tubes. However, it is likely that children seen in primary care clinics with OME have more variability in effusion characteristics before meeting the clinical indications for surgery, and methods to quantify these characteristics at earlier stages of OME would be of value to assess how these characteristics may also influence both WAI and other audiologic outcomes. Another limitation of this work is that four of the ears included in the study had a history of previous tube placement, and recent work from Hunter et al (2020) demonstrated an effect of a history of tube placement on ambient absorbance. However, a history of tube placement was demonstrated to cause an increase in ambient absorbance in the mid-frequencies, so given this effect is opposite to the effect of the effusion observed here, the small number of ears impacted, and the fact that the ears were split up within two effusion groups, it is unlikely a history of tube placement is a confounding factor here.

CONCLUSIONS

The goal of this work was to determine whether there is a systematic effect of middle ear effusion volume on WAI (absorbance) in children with surgically confirmed OME. Overall, our findings indicate that absorbance is a strong predictor of the volume of a middle ear effusion in children with OME. The presence of effusion, as well as the volume of a present effusion, were predicted based on absorbance responses at TPP with high sensitivity and specificity. Given associations between effusion volume and audiometric thresholds, these results suggest that WAI could provide objective insight into the hearing status of children with OME.

ACKNOWLEDGMENTS

We thank Lauren Crowther, Hannah Johnson, and Kayla Samuelson for assistance testing participants. We thank Dr. Kelli Rudman, Nicholas Warren, Jessalyn Bentz, Ashley Moore, Wendy Lyster, and ENT clinical and surgical staff for assistance in recruitment and data collection coordination, with Dr. Rudman also contributing effusion assessments at the time of surgery. We thank Dr. Emily Buss for the thoughtful conversations surrounding this work. We thank Dr. Adam Bosen for constructive input regarding the analysis approach. We thank Dr. Dominic Cosgrove and Dan Meehan for access to wet lab space and assistance with middle ear effusion analyses. Finally, we thank all of the participants and families who made this work possible.

REFERENCES

Alaerts J., Luts H., Wouters J. Evaluation of middle ear function in young children. Otol Neurotol, (2007). 28, 727–732.
Allen J. B., Jeng P. S., Levitt H. Evaluation of human middle ear function via an acoustic power assessment. J Rehabil Res Dev, (2005). 42(4 Suppl 2), 63–78.
Al-Salim S., Tempero R. M., Johnson H., Merchant G.R. Audiologic profiles of children with otitis media with efusion. Ear Hear, (2001). Mar. 26. [Epub ahead of print].
Anwar K., Khan S., Rehman H. U., Javaid M., Shahabi I. Otitis media with effusion: Accuracy of tympanometry in detecting fluid in the middle ears of children at myringotomies. Pak J Med Sci, (2016). 32, 466–470.
Baldwin M. Choice of probe tone and classification of trace patterns in tympanometry undertaken in early infancy. Int J Audiol, (2006). 45, 417–427.
Beers A. N., Shahnaz N., Westerberg B. D., Kozak F. K. Wideband reflectance in normal Caucasian and Chinese school-aged children and in children with otitis media with effusion. Ear Hear, (2010). 31, 221–233.
Blomgren K., Pitkäranta A. Is it possible to diagnose acute otitis media accurately in primary health care? Fam Pract, (2003). 20, 524–527.
Burdiek L. M., Sun X. M. Effects of consecutive wideband tympanometry trials on energy absorbance measures of the middle ear. J Speech Lang Hear Res, (2014). 57, 1997–2004.
Chow A. H. C., Cai T., McPherson B., Yang F. Otitis media with effusion in children: Cross-frequency correlation in pure tone audiometry. PLoS One, (2019). 14, e0221405.
Daly K. A., Hunter L. L., Lindgren B. R., Margolis R., Giebink G. S. Chronic otitis media with effusion sequelae in children treated with tubes. Arch Otolaryngol Head Neck Surg, (2003). 129, 517–522.
Di Francesco R. C., Barros V. B., Ramos R. Otitis media with effusion in children younger than 1 year. Revista Paulista de Pediatria, (2016). 34, 148–153.
Ellison J. C., Gorga M., Cohn E., Fitzpatrick D., Sanford C. A., Keefe D. H. Wideband acoustic transfer functions predict middle-ear effusion. Laryngoscope, (2012). 122, 887–894.
Feeney M. P., Grant I. L., Marryott L. P. Wideband energy reflectance measurements in adults with middle-ear disorders. J Speech Lang Hear Res, (2003). 46, 901–911.
Feeney M. P., Grant I. L., Mills D. M. Wideband energy reflectance measurements of ossicular chain discontinuity and repair in human temporal bone. Ear Hear, (2009). 30, 391–400.
Gravel J. S., Roberts J. E., Roush J., Grose J., Besing J., Burchinal M., Neebe E., Wallace I. F., Zeisel S. Early otitis media with effusion, hearing loss, and auditory processes at school age. Ear Hear, (2006). 27, 353–368.
Gravel J. S., Wallace I. F. Effects of otitis media with effusion on hearing in the first 3 years of life. J Speech Lang Hear Res, (2000). 43, 631–644.
Groon K. A., Rasetshwane D. M., Kopun J. G., Gorga M. P., Neely S. T. Air-leak effects on ear-canal acoustic absorbance. Ear Hear, (2015). 36, 155–163.
Guan X., Seale T. W., Gan R. Z. Factors affecting sound energy absorbance in acute otitis media model of chinchilla. Hear Res, (2017). 350, 22–31.
Hartley D. E., Moore D. R. Effects of otitis media with effusion on auditory temporal resolution. Int J Pediatr Otorhinolaryngol, (2005). 69, 757–769.
Huang G. T., Rosowski J. J., Puria S., Peake W. T. Tests of some common assumptions of ear-canal acoustics in cats. J Acoust Soc Am, (2000). 108(3 Pt 1), 1147–1161.
Hunter L. L., Blankenship C. M., Lin L., Sloat N. T., Perdew A., Stewart H., Moore D. R. Peripheral auditory involvement in childhood listening difficulty. Ear Hear, (2020). 42, 29–41.
Hunter L. L., Margolis R. H., Giebink G. S. Identification of hearing loss in children with otitis media. Ann Otol Rhinol Laryngol Suppl, (1994). 163, 59–61.
Hunter L. L., Prieve B. A., Kei J., Sanford C. A. Pediatric applications of wideband acoustic immittance measures. Ear Hear, (2013). 34(Suppl 1), 36S–42S.
Hunter L. L., Tubaugh L., Jackson A., Propes S. Wideband middle ear power measurement in infants and children. J Am Acad Audiol, (2008). 19, 309–324.
Jerger J. Clinical experience with impedance audiometry. Arch Otolaryngol, (1970). 92, 311–324.
Keefe D. H., Levi E. Maturation of the middle and external ears: Acoustic power-based responses and reflectance tympanometry. Ear Hear, (1996). 17, 361–373.
Keefe D. H., Sanford C. A., Ellison J. C., Fitzpatrick D. F., Gorga M. P. Wideband aural acoustic absorbance predicts conductive hearing loss in children. Int J Audiol, (2012). 51, 880–891.
Keefe D. H., Simmons J. L. Energy transmittance predicts conductive hearing loss in older children and adults. J Acoust Soc Am, (2003). 114(6 Pt 1), 3217–3238.
Koivunen P., Uhari M., Laitakari K., Alho O. P., Luotonen J. Otoacoustic emissions and tympanometry in children with otitis media. Ear Hear, (2000). 21, 212–217.
Liu Y. W., Sanford C. A., Ellison J. C., Fitzpatrick D. F., Gorga M. P., Keefe D. H. Wideband absorbance tympanometry using pressure sweeps: System development and results on adults with normal hearing. J Acoust Soc Am, (2008). 124, 3708–3719.
Mandrekar J. N. Receiver operating characteristic curve in diagnostic test assessment. J Thorac Oncol, (2010). 5, 1315–1316.
Margolis R. H., Hunter L. L., Giebink G. S. Tympanometric evaluation of middle ear function in children with otitis media. Ann Otol Rhinol Laryngol Suppl, (1994). 163, 34–38.
Margolis R. H., Saly G. L., Keefe D. H. Wideband reflectance tympanometry in normal adults. J Acoust Soc Am, (1999). 106, 265–280.
Maw A. R., Bawden R. Tympanic membrane atrophy, scarring, atelectasis and attic retraction in persistent, untreated otitis media with effusion and following ventilation tube insertion. Int J Pediatr Otorhinolaryngol, (1994). 30, 189–204.
Merchant G. R., Merchant S. N., Rosowski J. J., Nakajima H. H. Controlled exploration of the effects of conductive hearing loss on wideband acoustic immittance in human cadaveric preparations. Hear Res, (2016). 341, 19–30.
Merchant G. R., Röösli C., Niesten M. E., Hamade M. A., Lee D. J., McKinnon M. L., Ulku C. H., Rosowski J. J., Merchant S. N., Nakajima H. H. Power reflectance as a screening tool for the diagnosis of superior semicircular canal dehiscence. Otol Neurotol, (2015). 36, 172–177.
Merchant G. R., Siegel J. H., Neely S. T., Rosowski J. J., Nakajima H. H. Effect of middle-ear pathology on high-frequency ear-canal reflectance measurements in the frequency and time domains. J Acoust Soc Am, (2019). 20, 529–552.
MRC Multi-centre Otitis Media Study Group. Risk factors for persistence of bilateral otitis media with effusion. Clin Otolaryngol Allied Sci, (2001). 26, 147.
Nakajima H. H., Pisano D. V., Roosli C., Hamade M. A., Merchant G. R., Mahfoud L., Halpin C. F., Rosowski J. J., Merchant S. N. Comparison of ear-canal reflectance and umbo velocity in patients with conductive hearing loss: A preliminary study. Ear Hear, (2012). 33, 35–43.
Nakajima H. H., Rosowski J. J., Shahnaz N., Voss S. E. Assessment of ear disorders using power reflectance. Ear Hear, (2013). 34(Suppl 1), 48S–53S.
Nørgaard K. R., Fernandez-Grande E., Laugesen S. Incorporating evanescent modes and flow losses into reference impedances in acoustic Thévenin calibration. J Acoust Soc Am, (2017). 142, 3013.
Pichichero M. E. Diagnostic accuracy of otitis media and tympanocentesis skills assessment among pediatricians. Eur J Clin Microbiol Infect Dis, (2003). 22, 519–524.
Pichichero M. E., Poole M. D. Comparison of performance by otolaryngologists, pediatricians, and general practitioners on an otoendoscopic diagnostic video examination. Int J Pediatr Otorhinolaryngol, (2005). 69, 361–366.
Prieve B. A., Feeney M. P., Stenfelt S., Shahnaz N. Prediction of conductive hearing loss using wideband acoustic immittance. Ear Hear, (2013). 34(Suppl 1), 54S–59S.
Ravicz M. E., Rosowski J. J., Merchant S. N. Mechanisms of hearing loss resulting from middle-ear fluid. Hear Res, (2004). 195, 103–130.
Rosenfeld R. M., Kay D. Natural history of untreated otitis media. Laryngoscope, (2003). 113, 1645–1657.
Rosenfeld R. M., Shin J. J., Schwartz S. R., Coggins R., Gagnon L., Hackell J. M., Hoelting D., Hunter L. L., Kummer A. W., Payne S. C., Poe D. S., Veling M., Vila P. M., Walsh S. A., Corrigan M. D. Clinical practice guideline: Otitis media with effusion (update). Otolaryngol Head Neck Surg, (2016). 154(1 Suppl), S1–S41.
Sanford C. A., Brockett J. E. Characteristics of wideband acoustic immittance in patients with middle-ear dysfunction. J Am Acad Audiol, (2014). 25, 425–440.
Sanford C. A., Hunter L. L., Feeney M. P., Nakajima H. H. Wideband acoustic immittance: Tympanometric measures. Ear Hear, (2013). 34(Suppl 1), 65S–71S.
Sassen M. L., van Aarem A., Grote J. J. Validity of tympanometry in the diagnosis of middle ear effusion. Clin Otolaryngol Allied Sci, (1994). 19, 185–189.
Shahnaz N., Bork K., Polka L., Longridge N., Bell D., Westerberg B. D. Energy reflectance and tympanometry in normal and otosclerotic ears. Ear Hear, (2009a). 30, 219–233.
Shahnaz N., Longridge N., Bell D. Wideband energy reflectance patterns in preoperative and post-operative otosclerotic ears. Int J Audiol, (2009b). 48, 240–247.
Stinson M. R. Revision of estimates of acoustic energy reflectance at the human eardrum. J Acoust Soc Am, (1990). 88, 1773–1778.
Sun X. M. Wideband acoustic immittance: Normative study and test–retest reliability of tympanometric measurements in adults. J Speech Lang Hear Res, (2016). 59, 819–834.
Takeuchi K., Majima Y., Hirata K., Morishita A., Hattori M., Sakakura Y. Prognosis of secretory otitis media in relation to viscoelasticity of effusions in children. Ann Otol Rhinol Laryngol, (1989). 98, 443–446.
Vlachou S. G., Tsakanikos M., Douniadakis D., Apostolopoulos N. The change in the acoustic admittance phase angle: A study in children suffering from acute otitis media. Scand Audiol, (2001). 30, 24–29.
Voss S. E., Herrmann B. S., Horton N. J., Amadei E. A., Kujawa S. G. Reflectance measures from infant ears with normal hearing and transient conductive hearing loss. Ear Hear, (2016). 37, 560–571.
Voss S. E., Horton N. J., Woodbury R. R., Sheffield K. N. Sources of variability in reflectance measurements on normal cadaver ears. Ear Hear, (2008). 29, 651–665.
Voss S. E., Merchant G. R., Horton N. J. Effects of middle-ear disorders on power reflectance measured in cadaveric ear canals. Ear Hear, (2012). 33, 195–208.
Won J., Monroy G. L., Huang P. C., Hill M. C., Novak M. A., Porter R. G., Spillman D. R., Chaney E. J., Barkalifa R., Boppart S. A. Assessing the effect of middle ear effusions on wideband acoustic immittance using optical coherence tomography. Ear Hear, (2020). 41, 811–824.
Whitton J. P., Polley D. B. Evaluating the perceptual and pathophysiological consequences of auditory deprivation in early postnatal life: A comparison of basic and clinical studies. J Assoc Res Otolaryngol, (2011). 12, 535–547.
Keywords:

Absorbance; Otitis media; Otitis media with effusion; Wideband acoustic immittance; Wideband tympanometry

Copyright © 2021 Wolters Kluwer Health, Inc. All rights reserved