Background: The high rate of antibiotic prescriptions general practitioners (GPs) make for respiratory tract infections (RTI) are often explained by non-medical reasons e.g. an effort to meet patient expectations. Additionally, it is known that GPs to some extent believe in the necessity of antibiotic treatment in patients with assumed bacterial infections and therefore attempt to distinguish between viral and bacterial infections by history taking and physical examination. The influence of patient complaints and physical examination findings on GPs' prescribing behaviour was mostly investigated by indirect methods such as questionnaires.

Methods: Direct, structured observation during a winter "cough an cold period" in 30 (single handed) general practices. All 273 patients with symptoms of RTI (age above 14, median 37 years, 51% female) were included.

Results: The most frequent diagnoses were 'uncomplicated upper RTI/common cold' (43%) followed by 'bronchitis' (26%). On average, 1.8 (95%-confidence interval (CI): 1.7-2.0) medicines per patient were prescribed (cough-and-cold preparations in 88% of the patients, antibiotics in 49%). Medical predictors of antibiotic prescribing were pathological findings in physical examination such as coated tonsils (odds ratio (OR) 15.4, 95%-CI: 3.6-66.2) and unspecific symptoms like fatigue (OR 3.1, 95%-CI 1.4-6.7), fever (OR 2.2, 95%-CI: 1.1-4.5) and yellow sputum (OR 2.1, 95%-CI: 1.1-4.1). Analysed predictors explained 70% of the variance of antibiotic prescribing (R2 = 0,696). Efforts to reduce antibiotic prescribing, e.g. recommendations for self-medication, counselling on home remedies or delayed antibiotic prescribing were rare.

Conclusions: Patient complaints and pathological results in physical examination were strong predictors of antibiotic prescribing. Efforts to reduce antibiotic prescribing should account for GPs' beliefs in those (non evidence based) predictors. The method of direct observation was shown to be accepted both by patients and GPs and offered detailed insights into the GP-patient-interaction.