BACKGROUND: Glue ear (otitis media with effusion) is the most common reason for surgical intervention in children. AIM: To determine the yield and predictive value of a set of risk factors that predict persistence of glue ear over the interval from general practice referral to ear, nose and throat (ENT) consultation to ensure the appropriateness of referrals. DESIGN OF STUDY: Nested case control study SETTING: Sixteen ENT departments in the UK. METHOD: With the aid of audiometry and tympanometry, diagnostic information was collected on 548 children from 16 ENT departments after referral by their general practitioner (GP), as a lead-in to a clinical trial, the Trial of Alternative Regimens in Glue Ear Treatment (TARGET). Using cases and controls, children were classified as either having or not having persistent glue ear. Parental reports on an extensive list of risk factors were also collected. RESULTS: After adjustment for time waiting to be seen from GP referral and age at referral, four main significant factors emerged for persistence of glue ear. These were: referral between July and December (OR [odds ratio] = 1. 73, 95% CI = 1.15 to 2.6); having a mother who smokes ten or more cigarettes per day (OR = 1.7, 95% CI = 1.1 to 2.8); multiple upper airway symptoms (OR = 2.2, 95% CI = 1.5 to 3.2; and siblings with a history of glue ear (OR = 1.6 for one sibling versus none). CONCLUSION: For a child who is referred between July and December; who has two or more upper airway symptoms, who has a sibling who has had glue ear, and who has a mother who smokes ten or more cigarettes per day, the odds of having persistent glue ear are over ten times that of a child without adverse values on these factors.