The effect of a GP’s perception of a patient request for antibiotics on antibiotic prescribing for respiratory tract infections: secondary analysis of a point-prevalence audit survey in 18 European countries

Multivariable mixed effects logistic regression for the dependent variable antibiotic prescribing controlling for age, comorbidity, fever, perceived illness severity, illness duration, perceived request for antibiotics, country and strep A and/or C-reactive protein (CRP) point-of-care (POC) testing for all patients. Three important subgroups, patients diagnosed with a LRTI, throat infection, or ILI were further studied; relevant clinical examination findings, with a significant association in univariable analysis, were included. CRP testing was included for patients with a LRTI and strep A for patients with a throat infection. Practice was included as random effect. No interaction terms were added to the models. Reference categories were ‘no’ for POC testing, comorbidity, fever,, and perceived request, ‘mild’ for illness severity, and Belgium (lowest antibiotic prescribing proportion) for country. For age and duration of symptoms an increase of one unit was considered. ^aAdjusted odds ratios (OR) and 95% confidence intervals (CI); values significant at the 5% significance level in bold.