Otitis media with effusion (OME) provides an example of how ascertainment bias can induce spurious correlations. Early work suggested it impacted children's language, but when unbiased samples are studied, the effect is absent or very small
2. 2
Otitis media and language development
Illustration of ascertainment (sampling)
bias: a potential cause of spurious
associations
3. 3
Otitis media with effusion
• Middle ear fills with fluid
• Very common in young children
• Causes fluctuating conductive hearing loss
Healthy
tympanic
membrane
Chronic otitis
media
Otitis media with
tympanic membrane
hole or perforation
Otitis media
treatment -
tympanic membrane
hole or perforation
with a ventilator
tube in place
4. Fluid in middle ear affects speech perception
For many years thought unimportant because
hearing loss is mild and fluctuating
But in 1960s-1970s studies suggested OME was a
risk factor for language problems
x x
6. Child’s language status will determine
whether they get into clinical sample6
But problem of ascertainment bias
Holm & Kunze sample were children from ear-nose-throat
(ENT) clinic, most of whom had had surgery on their ears
OME is very common
Often treated conservatively, but…
• Parent more likely to go to doctor if child has
language problems
• Doctor more likely to refer to ENT if child has
language problems
• ENT surgeon more likely to operate if child has
language problems
7. The solution?
Need unbiased sample: study OME in a whole population
“Our results indicate no to very small negative
associations of OME and associated hearing loss
to children’s later speech and language
development.”
9. Always consider how people get into your
sample.
Ascertainment bias is a huge problem in
clinical studies – can induce apparent
correlations that are not present in the
whole population
The bottom line