Predicting bacterial cause in infectious conjunctivitis: Authors' reply

EDITOR—Glaser asked for clarification of “history of conjunctivitis.” We meant a self reported medical history. We asked patients if they had a history of allergic eye disease, prior infection of the eye with comparable symptoms, or other eye diseases.

We agree with Clarke that a conjunctivitis caused by Chlamydia can cause sticky eyes. A compromised immune system and an active sexually transmitted disease such as Chlamydia were exclusion criteria for our study. These conditions did not occur among the eligible patients.

Fierz's letter was mainly about steroids and keratitis. The issue of our study was an uncomplicated infectious conjunctivitis. Exclusion criteria were loss of vision, pain in the eye (eyeball), and wearing of contact lenses, all pertinent questions according to Fierz.

Davies would not use itching to make treatment decisions on the basis of our data. Still, itching fulfilled the criteria underlying our stepwise logistic regression approach. We entered variables whose association with a bacterial origin was fairly strong (univariate P value ≤ 0.1) into the model. We then excluded variables when in a multivariable model their corresponding P value exceeded 0.15. In a logistic model the association between variables, or the contribution of a variable taking into account the other variables, counts. In line with the statistical literature on predictive models, this approach explicitly allows entry of variables whose P value exceeds 0.05. In science, in exceptional cases only statistical significance is an issue. Our study was no such exception.

Davies was puzzled by the odds ratio of 2.68 for one glued eye. Davies thought that, according to the percentages in table 1, the odds ratio would be below 1.0. However, “glued eyes” was a determinant containing three categories, modelled as two dummies. The correct numbers of true positives, false positives, false negatives, and true negatives to calculate this odds ratio are 30, 74, 5, and 33, respectively.

We agree with Davies's comment on diagnosing ciliary redness over the telephone. Doctors should be very careful when diagnosing over the telephone. This should probably have been emphasised more in our paper.