Control of the temporal aspect when considering risk factors for acute otitis media.

A random sample of 2512 children was monitored to age 2 years to study the biologic effects of various risk variables on acute otitis media using a new dynamic modeling that controls both the confounding effects and time dependency. Dynamic modeling proved to be superior to conventional approaches, both the random and systematic error being much smaller and the effect estimates being biologically interpretable. The major risk factors were the existence of a previous episode of acute otitis media in general (odds ratio, 2.03; 95% confidence interval [Cl], 1.81 to 2.25) or particularly during the preceding 3 months (odds ratio, 3.74; 95% Cl, 3.40 to 4.10) and attending a day nursery (odds radio, 2.06; 95% Cl, 1.81 to 2.34). As the form of day care is the only modifiable risk variable of significant importance and previous episodes entail a risk of future ones, infants should be cared for at home, particularly after they have already experienced an episode of acute otitis media.