 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
* Department of Pediatrics, Division of Emergency Medicine
Southeast Child Safety Institute, Children's Hospital of Alabama, University of Alabama at Birmingham, Birmingham, Alabama
Departments of Pediatrics and Emergency Medicine, Vanderbilt University, Nashville, Tennessee
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Methods. A randomized, clinical trial was conducted among children 6 to 35 months of age who presented to a pediatric emergency department with either fever or upper respiratory infection symptoms. Children were randomized into 2 groups, in which the attending physician evaluated tympanometry results (Tymp Aware) or the attending physician was blinded to the tympanometry findings (Tymp Unaware). Tympanometry curves were graded independently by using a modified version of the Jerger scale.
Results. Of the 698 patients enrolled, tympanometry was performed successfully for 99.3%. Antibiotics were prescribed for OM for 27.9% of all patients. No statistically significant difference in antibiotic prescription rates for OM between the Tymp Aware group (28.8%) and the Tymp Unaware group (26.8%) was found. Of all patients for whom antibiotics were prescribed for OM, 14% had normal curves for both ears and 40% had some tympanographic movement bilaterally.
Conclusions. Tympanometry did not seem to change diagnoses or prescribing behavior in the group of physicians studied. Antibiotics were commonly prescribed for presumed OM in the absence of effusions documented with tympanometry.
Key Words: otitis media • tympanometry • antibiotic • antibiotic resistance • randomized trial design
Abbreviations: OM, otitis media • PEM, pediatric emergency medicine
An epidemic of microbial resistance is occurring in the United States, attributable in large part to overprescription of antibiotics.1,2 Although most upper respiratory infections are caused by viral agents, a large number of antibiotics are prescribed for management of those illnesses.3 Otitis media (OM) is the most prevalent upper respiratory condition diagnosed in the pediatric population, but there are limited objective criteria, such as tympanometry, available for accurate diagnosis.4 Acute OM is most commonly diagnosed among children 7 to 36 months of age; however, otoscopic examinations can be challenging to perform for patients in this age range.5
Tympanometry has been used in various forms for >40 years, and the feasibility of using this procedure in primary care settings has been established.6,7 Tympanometry measures the compliance of the tympanic membrane as pressure is varied in the ear canal; the method facilitates indirect determination of the presence or absence of fluid in the middle ear space. This modality provides an objective measure of ear canal volume, tympanometric peak pressures, gradients, and static admittance. A graphic curve is generated and can be compared with sample curves that represent various pathologic or normal conditions. Primary care physicians have demonstrated the ability to use this technique successfully among children with and without the diagnosis of acute OM.7
The Centers for Disease Control and Prevention and the American Academy of Pediatrics have recommended the use of diagnostic tools such as tympanometry, which can "aid in establishing the presence of fluid and in validating the examiner's skills through repeated use and comparison with visual observation."8(p167) Otoscopy alone is imprecise for the diagnosis of acute OM.9 As a consequence, clinicians prescribe antibiotics for nonbacterial conditions, a practice that likely contributes to the rapid progression of antibiotic resistance. This study was designed to test the following hypotheses: 1) the use of tympanometry may significantly change physician prescribing behavior, specifically reducing the use of antimicrobial agents; 2) with independently graded tympanometry results as a standard, physicians may overprescribe antibiotics for OM; and 3) tympanometry is an efficient modality that is acceptable to parents.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Eligibility Criteria
Patients were eligible for the study if they met the following criteria: 1) age between 6 and 36 months and 2) historical features of fever, rhinorrhea, cough, or any combination of those findings. Children were excluded if they were receiving antibiotics at the time of presentation, had a medical history of tympanic membrane surgery (including myringotomy tube placement), or had a history of craniofacial anomalies. Subjects with incomplete data cards and subjects for whom tympanometric examinations could not be performed were excluded from the final analysis.
Tympanometry Operators
The tympanometry operators were employed both to recruit patients for study and to perform tympanometry. Each operator received a minimum of 4 hours of tympanometry training with a pediatric audiologist and a minimum of 2 hours of consultation with the principal investigator (D.M.S), reviewing the study protocol and procedures. The tympanometry operators performed daily calibration checks before enrollment of patients, as instructed in the operating manual provided by the manufacturer.
Subjects meeting inclusion criteria were randomly assigned to 1 of 2 groups by using a computer-derived list of random numbers (True Epistat software, version 5.3; Epistat Services, Richardson, TX), after informed consent had been obtained from the parent or legal guardian of the child. The 2 groups were as follows: 1) Tymp Aware, in which tympanometry results were provided to the attending physician for analysis and use in patient care, and 2) Tymp Unaware, in which the attending physician was blinded to the tympanometry results. Tympanometry was performed for all subjects with a single frequency (226-Hz) tympanometer (Welch Allyn, Skaneateles Falls, NY). If tympanometry could not be performed, then the tympanometry operator indicated this on the patient data card and no additional data were entered. The operator recorded the following subject demographic information: 1) age (in months), 2) gender (male or female), 3) ethnicity (white, African American, Hispanic, or other), and 4) insurance status (Blue Cross-Blue Shield, Medicaid, other private plan, or self-pay). The tympanometry operator then rated the following 3 factors, using a 5-point Likert scale: 1) ease of obtaining tympanometry (1 = easy, 5 = difficult), 2) estimated time spent performing tympanometry (in minutes), and 3) perceived parental acceptance (1 = good, 5 = poor).
Physician Evaluation
All participating physicians had completed a minimum of 3 years of pediatric residency training. All study physicians received verbal training in tympanometry interpretation, accompanied by a written handout that included sample tympanometry curves provided by the manufacturer. The operator assigned all physicians a letter code, and the investigators were blinded to the code throughout the course of the study. The operator placed the letter code on the data sheet of each subject if tympanometry was successful.
All physicians were given a study data card to complete after tympanometry was performed. In the Tymp Aware group, physicians were provided with printed tympanometry results for each ear, labeled left or right. Attached to the chart for the Tymp Aware group was a tympanometry interpretive guide, describing sample curves, provided by the tympanometer manufacturer. Participating physicians had the option of reviewing the tympanogram before or after examination of the patient. In the Tymp Unaware group, the physician was blinded to the tympanometry results; these results were placed in the patient's folder after the assigned physician had completed the data card.
After completing a history and physical examination of the patient, including otoscopic examination, participating physicians circled 1 of 4 options for each ear (acute OM, serous OM, no OM, or unclear). For the choice of unclear, a space was provided for an explanation. All examination rooms were equipped with round otoscopic heads with an attached pneumatic insufflation device. For the Tymp Aware group only, the same 4 options as for the otoscopic findings were listed for the tympanometric evaluation. All physicians in both groups were required to circle the following: final diagnosis (acute OM, serous OM, no OM, or unclear) and treatment choice (antibiotic or no antibiotic). All physicians in both groups were required to answer the following 2 questions: "Was an antibiotic prescribed primarily for otitis media?" (yes or no) and "If antibiotic selected, which one chosen?" (amoxicillin, Augmentin, or other, with a space provided to specify the antibiotic prescribed). Finally, physicians in the Tymp Aware group circled a response to the following question: "Do you think the Tymp reading altered your treatment plan today?" (yes/no).
Independent Tympanometry Data Interpretation
All tympanometry curves were graded by an investigator (D.M.S) who was blinded to the final diagnosis. Ear canal volume of <0.2 cm3 was considered inadequate and was not used for interpretation. All curves were independently assigned to category A, B, or C on the basis of a modified version of the Jerger scale.10 An A curve was defined as a bell-shaped curve with tympanometric peak pressures in the range of –150 to +100 dPa, a B curve was defined as a flat line, and a C curve was any curve not meeting A or B criteria, as described. A pediatric audiologist who was blinded with respect to the final diagnosis and investigator interpretation reviewed 12% (N = 100) of all curves, to determine the accuracy of the investigator-interpreted tympanometry curves.
Statistical Analyses
To detect a 
10% difference in antibiotic prescriptions written between the 2 groups (Tymp Aware versus Tymp Unaware), using a 2-tailed 
= .05 and a power of .80, each group required a minimum of 220 patients for study. Median values were calculated for all Likert scale analyses. 
2 and t tests were used to compare descriptive data among the study groups and to evaluate outcomes of interest, specifically antibiotic prescriptions for OM and final diagnoses, between groups. The z test for proportions was used to evaluate tympanometry curves according to group designation. For all statistical testing, a P value of <.05 was considered significant.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
2 = 5.24, P = .02) were found in the Tymp Unaware group, compared with the Tymp Aware group (61.5% vs 52.5%). The median time to perform tympanometry was 1 minute for 85.5% of all children evaluated. Performing tympanometry was described by the operator as easy for most patients (85.9%), and perceived parental acceptance was described as good for >98.5% of all evaluated patients.
|
2 = .00, P = .96) or final diagnosis (2 = 1.68, P = .64). In addition, there was no significant difference between groups with respect to type of antibiotic prescribed (2 = 3.07, P = .69). The most common antibiotics specifically prescribed for OM for the 2 groups combined were amoxicillin (56.6%), "other antibiotic" (25.4%), and amoxicillin clavulanate (18.0%). The Tymp Aware group use of antibiotics decreased significantly with time (2 = 9.53, P = .05). The Tymp Unaware group use of antibiotics did not change significantly with time (2 = 4.25, P = .37).
|
2 = 1.24, P = .27). The proportions of PEM versus non-PEM physicians were similar in the 2 groups (2 = 0.43, P = .51). In 2.8% of the Tymp Aware encounters, physicians indicated that tympanometry altered the treatment plan. Table 3 describes retrospectively reviewed tympanometry curves for patients who were prescribed antibiotics specifically because of the diagnosis of OM. There were no significant differences between the Tymp Aware and Tymp Unaware groups for the 3 tympanogram categories, ie, 1) normal curves bilaterally, 2) some curve movement bilaterally, or 3) flat curve on either side. Of the group of children for whom antibiotics were prescribed for OM (N = 189), 40.6% had some tympanometric movement (either type A or C) in both ears, whereas 14.3% had normal (type A) curves bilaterally. The overall rate of concordance between physician- and audiologist-interpreted curves was 85.5%. The subset A, B, and C curve concordance rates were 93.8%, 98.1%, and 69.7%, respectively. Of the discordant C curves, 70% were interpreted as a C curve by the physician reviewer and as an A curve by the audiologist.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Myringotomy, the traditional standard method, was used in previous studies to determine diagnostic outcome differences using tympanometry, as opposed to specific treatment outcomes such as antibiotic prescription rates.6,15 Finitzo et al16 studied pneumatic otoscopy and tympanometry with findings at myringotomy. With respect to diagnostic accuracy, pneumatic otoscopy alone revealed a specificity of 58%. Compared with otoscopy, tympanometry was significantly more specific in determining the noneffusion state. Combining tympanometric and otoscopic examinations yielded a specificity of 93% in determining the presence of myringotomy-confirmed middle ear effusions. Other reports also demonstrated that tympanometry improves the diagnostic accuracy for acute OM, but few primary care clinicians use this objective, inexpensive, diagnostic measure on a routine basis.15,17
To our knowledge, this is the first prospective trial in any age group using tympanometry as the randomized variable. Our results indicate that the group of pediatricians studied did not alter their final diagnoses or prescribing habits on the basis of tympanometry results. Johansen et al18 evaluated whether tympanometry altered postotoscopic diagnoses by physicians for children 
16 years of age. Nearly one-quarter of all providers changed their initial otoscopic diagnoses after being provided with the results of tympanometry. Similar to our experience, the frequency of antibiotic prescriptions in that study was unaffected.
A substantial number of children in our study group were prescribed antibiotics for OM despite having either normal tympanograms (type A) bilaterally or significant movement in the tympanometry curves (type A or C) bilaterally. The presence of a type A curve significantly reduces the likelihood of a middle ear effusion.15,19 Therefore, a total of 14% of antibiotics might have been prescribed unnecessarily for OM in this analysis, which is consistent with previous reports documenting, with myringotomy, false-positive otoscopic diagnoses in the range of 9% to 14%.15,17 Although the Tymp Aware group exhibited decreased antibiotic usage rates with time, we think this is likely the result of statistical imprecision attributable to a small sample size in the first quarter.
An additional 26% of curves were retrospectively interpreted as type C curves bilaterally or some combination of C and A curves for both ears. A recent report by Palmu et al,20 investigated type C negative-pressure tympanograms in a large cohort of children <24 months of age. In that study, myringotomies were performed when pneumatic otoscopy suggested OM. A small percentage (15%) of children with negative tympanometric peak pressures were found to have fluid in the middle ear space, with a majority of the effusion-positive middle ear aspirates being nonbacterial in nature. These findings suggest that many type C curves may be more indicative of eustachian tube dysfunction than bacterial OM.
There are several limitations to this study. Despite randomization, there was a significant gender difference between the 2 groups. However, there were no statistical gender-related differences with respect to the 2 main outcomes of interest, ie prescription of antibiotics for OM and final diagnoses. The study design was unable to control for the temporal relationship between tympanometry and otoscopy. It is possible that controlling this aspect of the study might have produced different results. Although pneumatic otoscopy has been demonstrated to improve clinicians' ability to diagnose OM accurately, rates of pneumatic insufflation were not recorded as a study variable and might have affected physician decision-making. Physicians were not encouraged in or discouraged from using the pneumatic otoscope, nor were they queried regarding whether they used this method. The study was designed to evaluate the specific effect of tympanometry on the clinicians' standard method of practice. In addition, the independent tympanometry review did not specifically designate C curve subtypes, which have been described as representing differing clinical findings in various reports.16,20 Although our retrospective interpretation of type C curves was unique, there was a high rate of agreement between the principal physician reviewer and the audiologist with the proposed A, B, and C categories, modified from the Jerger classification system. The high rate of concordance between the physician findings and the audiologist findings confirms the ability of outpatient-based physicians to interpret tympanograms accurately.7
We hypothesize that the physicians in the Tymp Aware group did not alter diagnosis or treatment decisions on the basis of the provided tympanographic information. If tympanometry had been used to exclude the absence of a middle ear effusion, our data indicate that 14% to 40% of OM diagnoses might have been avoided. Our findings are similar to the findings of Pichichero,9 who found that US physicians overdiagnosed acute OM with a mean frequency of 26%. Tympanometry was introduced as a new modality during the course of study in our emergency department, which might have influenced the participating physicians to favor their otoscopic examination findings over the tympanometry results. Physicians were intentionally trained with brief verbal and written instructions provided by the manufacturer. This more realistically simulated the introduction of this device for providers new to its use. An alternative approach might include physician tympanometry education provided during the period of tympanogram interpretation. Such point-of-care educational intervention has been demonstrated to significantly change physician prescribing behavior.21 We hypothesize that, if physicians used tympanometry to improve diagnostic specificity, this modality would substantially reduce unnecessary antimicrobial treatment.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
We thank all of the faculty members and fellows who participated in the study in the Department of Pediatrics, Division of Emergency Medicine, University of Alabama at Birmingham. Special thanks go to Heather Wigley Baty, MCD, MBA, CCC-A, who kindly donated her audiology training expertise and support for the project, and to M. Douglas Baker, MD, and Eugene Shapiro, MD, who carefully reviewed the manuscript.
|
FOOTNOTES |
|---|
Reprint requests to (D.M.S.) Pediatric Emergency Department, Yale-New Haven Children's Hospital, 20 York Street, Room WP-143, New Haven, CT 06504. E-mail: david.spiro{at}yale.edu
Dr Spiro is currently affiliated with the Section of Emergency Medicine, Department of Pediatrics, Yale University School of Medicine.
This research was presented at the Pediatric Academic Societies Meeting, May 3-6, 2003.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  S. B. Meropol, H. A. Glick, and D. A. Asch Age Inconsistency in the American Academy of Pediatrics Guidelines for Acute Otitis Media Pediatrics, April 1, 2008; 121(4): 657 - 668. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Other articles noted Evid. Based Med., November 1, 2004; 9(6): 191 - 192. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2004 American Academy of Pediatrics. All rights reserved. |
||
|
||
