Read more

November 04, 2024
4 min read
Save

Q&A: App ‘much better’ than physicians at diagnosing ear infections

You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

Key takeaways:

  • Among children with ear infections who had S. pneumoniae isolated from middle ear fluid, half had serotypes in PCV13.
  • Researchers used a smartphone app to confirm whether or not patients had ear infections.

More than half of children with acute otitis media and 32% of children with upper respiratory infections tested positive for multiple pathogens in their nasal passages, according to a study presented at IDWeek.

“Despite changes and updates over time in pneumococcal vaccines that we are administering to children, we continue to observe both Haemophilus influenzae and Streptococcus pneumoniae in specimens of middle ear fluid in children less than 3 years of age with acute otitis media (AOM),” Judith M. Martin, MD, professor of pediatrics at the University of Pittsburgh School of Medicine, told Healio about the study, which was selected as the Pediatric Infectious Diseases Society featured abstract.

IDC1024Hoberman_graphic


Vaccine serotypes still infecting children

Martin and colleagues analyzed middle ear fluid samples and nasal samples from 72 children with AOM and nasal samples only for 322 children with AOM and 151 with upper respiratory infections. All children had visited a primary care or ear, nose and throat practice in Pittsburgh between October 2019 and January 2024.

Out of all 545 children aged 6 to 35 months, 44% had S. pneumoniae, 37% had H. influenzae, and 66% had Moraxella catarrhalis. More than half tested positive for multiple pathogens in their nose. The researchers noted that having multiple pathogens was more common among children with AOM than children with upper respiratory infections (56% vs. 32%).

Martin told Healio that the pneumococcal serotypes present in children with AOM shift over time in response to vaccine coverage.

“This is called serotype replacement,” Martin said. “The vaccines include the serotypes that cause invasive disease and those serotypes that are more likely to be penicillin nonsusceptible.”

Martin said invasive disease has decreased dramatically since pneumococcal vaccines became routine in 2000.

However, there are some serotypes covered by vaccines that are still infecting children. Among children who had S. pneumoniae in the nose, the most common serotypes were 35B, 15B, 3 and 15C. Serotypes identified in middle ear fluid included 3, 19A, 11A and 15B. Of these, half were serotypes that are included in the PCV13 vaccine. Among children with S. pneumoniae middle ear fluid isolates, 90% had identical serotypes in a nasal specimen.

“We are still observing serotypes that are in the current vaccines in the specimens of middle ear fluid that we tested,” Martin said. “These include serotypes such as serotype 3, which we know that the current vaccines are not providing great protection against.”

‘Much better than most of us pediatricians’

Healio Pediatrics Peer Perspective Board Member Alejandro Hoberman, MD, chief of the division of general academic pediatrics at the University of Pittsburgh School of Medicine, said diagnosing AOM in the first place is a struggle for many physicians.

In a previous study, Hoberman and colleagues found that most primary care and pediatric providers are less than 75% accurate at diagnosing ear infections.

Hoberman and colleagues have designed a phone app to help clinicians diagnose AOM, which outperformed physicians — including pediatricians — in a study published earlier this year. We spoke with Hoberman about the app, which is not publicly available yet.

Healio: Why do physicians struggle to diagnose ear infections?

Hoberman: You have a wiggly target, a baby in the parents’ arms or on the examining table who is moving. There is usually cerumen obstructing the view of the otoscope, and you usually get a quick glimpse of the eardrum. Once you are looking at the eardrum, you must properly diagnose bulging of the tympanic membrane, which is the hallmark of acute otitis media. Anything less than bulging of the eardrum is not going to be classified as an acute otitis media, and those patients should not be treated with antibiotics.

It is an art, and the more years you have, and the more sophisticated training, a higher accuracy is achieved. But having said so, the accuracy varies between 40% or 50% to 80%, so we are not very accurate at diagnosing acute otitis media.

Healio: What are the consequences of misdiagnosing ear infections?

Hoberman: If it is missed, a misdiagnosis will result in undertreatment and potentially allowing that child to go for another day or a couple of days with ear pain that is going to be untreated. Overdiagnosis results in unnecessary use of antibiotics, which increases the likelihood of resistance, or selection of strains that are more resistant. Overuse of amoxicillin may have resulted in a higher number of strains of Haemophilus influenzae that produce beta lactamase. Both overdiagnosis and underdiagnosis are detrimental.

Healio: Tell me about the app you designed to aid in diagnosing AOM.

Hoberman: We trained a deep residual-recurrent neural network on videos of the tympanic membranes that we captured using an otoscope to classify them into acute otitis media or no acute otitis media. It did not discriminate between otitis media with effusion and a normal tympanic membrane because that does not change your management.

We trained two advanced programs. The first one was a quality filter that allows us to determine if the video of the tympanic membrane is good enough to interpret for algorithmic purposes. For example, can you see the eardrum? That one ruled out 90% of the inappropriate images. If the images were decent, we would run the classifier. All we needed was a 3-second video that allowed us to identify it in two different ways: one was using the deep residual neural network, and the other one was using a decision tree model. Both gave very identical results that were 94% accurate, which is much better than most of us pediatricians, family docs and folks are at diagnosing ear infections. We tested it on 1,151 eardrums, and now we are working through a commercial partner to actually try to bring this to the market so it can be widely used.

The app is available for many of our offices in Pittsburgh that are participating in clinical research — we used it in the studies that we are doing on pathogens that are causing acute otitis media. We are using it for validation purposes to be sure that every child we are enrolling in the study truly has AOM. We are using it in the clinical context in our primary care center here, but it is not available for people to download and use yet.

Healio: How will this app help providers and patients?

Hoberman: There are two benefits of this tool, not only do we get a more accurate diagnosis, but we can show parents the image of the tympanic membrane that we just obtained and explain why this child does or does not need antibiotics. It also allows us to train our medical students and residents by looking at the same thing that we are looking at.

For more information:

References: