Figure 1: Diagram showing gold standard assessment and three
further strategies for diagnosis: tele-diagnosis, non-specialist
diagnosis and prediction model diagnosis.
To date few studies have explored the validity or feasibility of such
approaches for diagnosis of external and middle ear disease.
Tele-diagnosis using otoscopic images captured with an endoscope or
smartphone has been shown to be successful if users are trained in image
capture,5 but appears unsuccessful if users are given
only instructions on use.6,7 The requirement for
training may therefore limit the scalability of this approach.
Studies of non-expert diagnosis of outer and middle ear disorders are
sparse, but results to date are not encouraging. In high income
settings, medical students, audiology students and general practitioners
report low self-confidence in diagnostic otoscopy.8 A
recent study found poor reliability of nurse otoscopic diagnosis in
Malawi.9 Validity of non-expert diagnosis therefore
remains uncertain.
Prediction model diagnosis is largely unexplored in this field. For
example (and to our knowledge) no study has explored if an algorithm
using patient history is valid to diagnose or differentiate external or
middle ear disorders.
Here we devised a novel study to directly compare the utility of
tele-diagnosis, non-expert diagnosis, and prediction model diagnosis as
a screening tool for common external and middle ear disorders. Our test
bed was a cohort of patients presenting to an Ear, Nose and Throat (ENT)
clinic in [removed for blind peer review] hospital in Cambodia
(www.csc.org), a setting with high prevalence
of otological disease.