Detection of middle-ear fluid in children with spectral gradient acoustic reflectometry: A screening tool for nurses?

Abstract

Objective. To evaluate the validity of spectral gradient acoustic reflectometry (SG-AR) in the hands of nurses in screening children for middle-ear fluid (MEF). Design. Prospective, blinded study. Setting. A satellite study within the Finnish Otitis Media Vaccine Trial in primary care in 1995–99. Patients. Some 739 ear examinations among a cohort of 271 children under the age of 2 years during different healthcare contacts (acute sick visits, check-up visits after otitis media, and scheduled healthy control visits at 24 months of age). Main outcome measures. Specificity, sensitivity and positive and negative predictive values of SG-AR performed by nurses in detecting MEF using pneumatic otoscopy by trained physicians as a reference. Results. SG-AR was successful in 585 (79%) ears. None of the cut-off points assessed resulted in both excellent sensitivity and specificity. Nevertheless, at the sick visits, positive predictive values at 50 and 60 degrees were as high as 94% and 89%, respectively. However, negative predictive value for the cut-off point of 100 degrees was considerably lower, at 75%. At the non-acute visits, negative predictive values were excellent. Conclusion. SG-AR is a useful device for nurses in screening MEF among children. It detects ears with both very high and very low probability of MEF and is especially effective in detecting MEF during sick visits and in ruling out MEF among non-acute patients. We recommend the use of the SG-AR cut-off point of 60 degrees as a sign of MEF when screening sick children, and the cut-off point of 100 degrees as a sign of a healthy ear among non-symptomatic patients.

• Diagnostic evaluation
• family practice
• middle ear infection
• screening

In a Finnish cohort study, the cumulative proportion of children who experienced at least one episode of acute otitis media (AOM) was 71% by the age of two years [1]. The reliability of the diagnosis is important, not only because of this high incidence of AOM but also because of increased antibiotic resistance. Diagnosis of AOM is based on the presence of fluid in the middle ear. Among methods to identify middle-ear fluid (MEF), myringotomy is invasive and pneumatic otoscopy is considered difficult to perform and is highly dependent on the cooperation of the child [2]. Tympanometry yields sensitivities and specificities comparable to those of pneumatic otoscopy in detecting MEF [3–5], but has the disadvantage of requiring an airtight seal in the ear canal and the results are thus affected by the cooperation of the child [3]. Moreover, the tympanogram is not always easy to interpret [6], even though with instruction and training good interobserver agreement can be achieved [7].

Acoustic reflectometry determines the probability of MEF by measuring the response of the tympanic membrane to a sound frequency spectrum sweep [8]. Spectral gradient acoustic reflectometry (SG-AR) analyses and displays the result graphically as a curve, from which the spectral gradient angles can be assessed [9]. MEF restricts the movement of the tympanic membrane and narrows the angle. SG-AR has shown to be comparable to pneumatic otoscopy and tympanometry in detecting MEF in studies, where the middle ear status has been confirmed by either myringotomy or pneumatic otoscopy [2], [8], [10–12]. Acoustic reflectometry does not require an airtight fit in the ear canal and is hence generally reliable even with struggling or crying children [2].

A reliable screening method for detection of middle-ear fluid, which can be handled by nurses, would be useful in primary care.

• Spectral gradient acoustic reflectometry performed by nurses was useful among children for detecting otitis media during acute sick visits.

• The method could be used for ruling out middle-ear fluid when screening non-symptomatic subjects.

This study evaluated SG-AR performed by a nurse as a screening tool to detect MEF in an outpatient setting. Screening of MEF among the vast number of children with upper respiratory infection symptoms and ear check-ups after otitis would save physicians’ time if otitis could be reliably ruled out by nurses. This would decrease the waiting time in the outpatient clinics by patients and their families and, ultimately, might help to reduce the pressure for unnecessary antibiotic prescriptions with the aid of a restrictive prescription policy and parent counselling [13].

Material and methods

The present study was a satellite study within the Finnish Otitis Media Vaccine Trial [14] conducted in 1995–99 among 2497 children enrolled at 2 months of age from three Finnish municipalities. All visits of a sub-cohort of 271 consecutive children in two of the main study sites were included in this study between August 1998 and March 1999. Patient characteristics and study visits are presented in Table I.

Table I.  Patient characteristics (n = 271) and study visits.

Boys	48%
Number of visits per child	1–6
Number of ear examinations	739 among 271 children (2.7 per child)
Number of successful ear examinations^1	585 among 246 children (2.4 per child)
Ear examinations at different healthcare visits
Acute sick visit	230 (39%)
Control 3–5 weeks after acute otitis media	90 (15%)
Scheduled healthy follow-up at 24 months of age	265 (45%)

Mean age (range)

23 months (1.5–2.0 years)

¹Technically successful SG-AR recordings among the 739 original recordings.

Exclusion criteria included a draining ear and a patent ventilation tube or other perforation in the tympanic membrane, after which the number of ear examinations performed in the study was 739 (mean 2.7 per child). EarCheck PRO® instrument with a SmartBase® recorder/printer unit (MDI Instruments Inc., Woburn, MA, USA) was used. A tutorial on the procedure of two hours was given to seven study nurses, which they performed according to the manufacturer's instructions in conjunction with the normal interview before the trial visits in sitting position. The instrument displays an error symbol when it detects sounds outside the expected range (e.g. if the child cries exceptionally loud) or cannot obtain a reading (e.g. if the ear canal is blocked or the tympanic membrane is perforated). In the case of “error” in the first SG-AR, the ear was examined up to three times. If all three attempts produced an error sign, the examination was considered to be failed. Immediately after SG-AR reading, one of nine study physicians performed pneumatic otoscopy, the criteria for clinical suspicion of otitis media being abnormal tympanic membrane finding (colour, position, or mobility) suggesting MEF. At the acute sick visits, myringotomy – if indicated by the clinical suspicion – was performed under local anaesthesia, and the presence of MEF was verified with suction. The setting was blinded: the nurse performing the SG-AR had no knowledge of the middle-ear status, and the physician performing the pneumatic otoscopy was unaware of the SG-AR results. The SG-AR results obtained by nurses were compared with physicians’ findings considered as reference standard (pneumatic otoscopy with or without myringotomy). With every SG-AR angle degree cut-off point, values under the cut-off point indicated a positive result (MEF), and values above the cut-off point a negative result (healthy ear). The sensitivity (ability to detect MEF and avoid false negatives), specificity (ability to detect healthy ears and avoid false positives), and positive (probability of MEF in the case of a positive test result) and negative predictive value (probability of healthy ear in the case of a negative test result) of SG-AR were calculated with their respective 95% confidence intervals within the different clinical contexts (i.e. during different types of healthcare visits).

Figure 1.  The spectral gradient acoustic reflectometry device used in this study, EarCheck PRO® instrument with a SmartBase® recorder/printer unit (MDI Instruments Inc., Woburn, MA, USA).

The study protocol was approved by the Ethical Review Board of National Public Health Institute of Finland and by the administrative boards of the municipalities, from where the study population was enrolled. Informed consent was obtained from the parents. The results are reported according to the STARD recommendations for diagnostic accuracy studies [15].

Results

SG-AR was unsuccessful in 154 ears (21%), making the final number of recorded ear examinations 585. Among them, no protocol deviations took place. MEF was found in 145 ears (25%) by the study physicians using pneumatic otoscopy. The presence of MEF was verified by a positive myringotomy finding in 94/145 (65%) of the MEF cases. At the sick visits, in 136/230 (59%) of the cases myringotomy was not performed, in 29 (13%) because of parental refusal and in 107 (46%) because of normal otoscopy finding. Sensitivities and specificities of SG-AR in detecting MEF at different spectral gradient cut-off points are presented for sick visits (Table II) and for non-acute visits (scheduled follow-ups and check-ups after AOM) (Table III) with their respective positive and negative predictive values and likelihood ratios. None of the individual cut-off points resulted in both high sensitivity and specificity. However, some clinically useful cut-off points were found.

Table II.  Sensitivity, specificity, positive (PPV) and negative (NPV) predictive values, and positive (LR + ) and negative (LR − ) likelihood ratios of spectral gradient acoustic reflectometry performed by nurses in detecting middle-ear fluid during unscheduled acute sick visits (n = 230 ears, prevalence of middle-ear fluid 50%).

Spectral gradient (degrees)	Sensitivity (%) (95% CI)	Specificity (%) (95% CI)	PPV (%) (95% CI)	NPV (%) (95% CI)	LR +	LR −
<50 (n = 33, 14%)	27 (19–35)	98 (96–100)	94 (86–100)	58 (51–65)	15.8	0.7
<60 (n = 54, 23%)	42 (33–51)	95 (91–99)	89 (81–97)	63 (55–70)	8.1	0.6
<70 (n = 85, 37%)	54 (45–64)	80 (73–87)	73 (64–82)	64 (56–72)	2.7	0.6
<80 (n = 118, 51%)	69 (61–78)	66 (58–75)	67 (58–75)	69 (60–77)	2.1	0.5
<90 (n = 141, 61%)	77 (69–85)	54 (45–63)	62 (54–70)	71 (61–80)	1.7	0.4
<100 (n = 169, 73%)	87 (81–93)	40 (31–49)	59 (51–66)	75 (64–86)	1.4	0.3

CI = confidence interval.

Table III.  Sensitivity, specificity, positive (PPV) and negative (NPV) predictive values, and positive (LR + ) and negative (LR − ) likelihood ratios of spectral gradient acoustic reflectometry performed by nurses in detecting middle ear fluid at non-acute trial visits¹ (n = 355 ears, prevalence of MEF 9%).

Spectral gradient (degrees)	Sensitivity (%) (95% CI)	Specificity (%) (95% CI)	PPV (%) (95% CI)	NPV (%) (95% CI)	LR +	LR −
<50 (n = 8, 2%)	19 (5–33)	99 (99–100)	75 (45–100)	93 (90–96)	31.3	0.8
<60 (n = 22, 6%)	42 (25–59)	97 (95–99)	59 (39–80)	95 (92–97)	15.1	0.6
<70 (n = 51, 14%)	48 (31–66)	89 (88–92)	30 (17–42)	95 (92–97)	4.4	0.6
<80 (n = 86, 24%)	61 (44–78)	79 (75–84)	22 (13–31)	96 (93–98)	3.0	0.5
<90 (n = 129, 36%)	71 (55–87)	67 (62–72)	17 (11–24)	96 (93–100)	2.1	0.4
<100 (n = 186, 52%)	87 (75–99)	51 (45–56)	15 (9–20)	98 (95–100)	1.8	0.3

¹Control check-up visits 3–5 weeks after otitis media (n = 90) and scheduled trial healthy follow-up visits at 24 months of age (n = 265). CI = confidence interval.

The prevalence of MEF varied significantly between the different types of visits. At sick visits MEF was present in 114/230 ears (50%), whereas at control follow-ups 3–5 weeks after otitis media MEF was found in 18/90 ears (20%), and at 24 months scheduled follow-up visit in 13/265 ears (5%). The sensitivities of SG-AR were essentially similar, but the specificities somewhat lower at the sick visits (see Tables II and III). Due to the varying prevalence of MEF in different clinical contexts (and, thus, varying pre-test probability of MEF), the positive and negative predictive values of SG-AR were clearly different at the sick visits: positive predictive values were higher (94% and 89% with cut-off points of 50 and 60 degrees, respectively) and negative predictive values lower (Tables II and III). The negative predictive value for the SG-AR value of 100 degrees was excellent (98%) for ears examined at non-acute visits.

Discussion

In our study, the rate of unsuccessful SG-AR recordings was higher and the diagnostic accuracy lower than in studies where SG-ARs had been performed by a physician [2], [8], [10], [11]. This is probably in part due to the younger patient material in this study (mean age 23 months) – a fact that also makes the present study more clinically relevant, since it reflects the child population with frequent AOM. Myringotomies performed in this study might have contributed to the lower success rate by causing additional fear of ear examinations. Still, SG-AR performed by nurses reached good positive and negative predictive values as a detector of MEF when pneumatic otoscopy by a physician was used as a comparison. This suggests that its use does not require much experience or expertise; only a short tutorial seemed to suffice. However, there were certain limitations. SG-AR seemed to both confirm and rule out MEF reliably at the ends of the SG-AR degree spectrum. Among all patients, 270 examinations (46%) were within these limits: 41 (7%) were “certain” MEF (<50 degrees) and 230 (39%) “certain” healthy ears (≥100 degrees). Results in the middle warrant, in our opinion, a check-up by a physician. Also the positive SG-AR finding by a nurse should be confirmed visually by a physician to rule out complications of AOM and to discriminate infectious and non-infectious MEF.

It is noteworthy that the positive and negative predictive values of SG-AR in detecting MEF were clearly different at different types of healthcare contacts because of varying prevalence of MEF. At sick visits, when the probability of MEF is higher than during scheduled check-ups, the positive predictive value was higher and the negative predictive value lower than among other groups. Also, at sick visits the percentage of test-positive SG-ARs (<50 degrees) was significantly higher than among the whole study population (14% vs. 7%). This improves the usability of SG-AR as a detector of MEF in the normal primary care setting with sick children. With a slightly slacker screening cut-off point (60 degrees), the positive predictive value was still excellent (89%) and the percentage of test positive ears increased to 23% among patients at the sick visit.

According to the results of this study, if the reason for the patient visit is the suspicion of MEF in a child with no acute upper respiratory symptoms (for example, a regular check-up after AOM), patients with negative SG-AR finding could be sent home without consulting a physician, since the negative predictive values were extremely high. This would save the time of healthcare professionals and the families, and economic resources. At the sick visits, however, the negative predictive values were considerably lower (<75%) and, thus, a significant proportion of the patients (i.e. 25%) with a negative SG-AR finding actually had MEF. Thus, SG-AR in the hands of nurses does not seem to be accurate enough to rule out otitis media in high-prevalence circumstances.

The limitation of this study is the use of pneumatic otoscopy instead of microscopic myringotomy with suction as the reference for determining the presence of MEF. However, the otoscopists were full-time study physicians, had clearly established criteria for MEF, and performed myringotomy with aspiration for a large proportion of subjects with findings suggestive of MEF; therefore their assessment can be considered reliable. The test reproducibility (inter-observer validity between nurses) was not assessed, which can be considered a limitation of the study.

We found the SG-AR to be a useful procedure, although the success rate of recordings was not optimal. However, acceptable sensitivities, specificities, and positive and negative predictive values were reached in detecting MEF in the hands of nurses relatively unfamiliar with the device, suggesting its usability in primary care as a screening tool. It was especially efficient in distinguishing the patients with both very high and very low probability of MEF. To clarify the role of SG-AR as a screening device, studies comparing acoustic reflectometry and tympanometry by nurses are warranted, since the latter is becoming increasingly popular in Finland. In the future, SG-AR could also be eligible as a home-screening tool, as some studies have already suggested [11], [16]. This, however, might compromise the objectivity of result interpretation. We recommend the use of the SG-AR cut-off point of 60 degrees as a sign of MEF when screening sick children, and the cut-off point of 100 degrees as a sign of a healthy ear among a non-symptomatic population. The latter finding does not need to be confirmed by a physician.

The author would like to thank the study nurses and physicians and the children and families participating in the study. The FinOM Studies were supported by Merck, AventisPasteur, and Wyeth-Lederle.