Delayed Antibiotic Prescription by General Practitioners in the UK: A Stated-Choice Study

Abstract

Delayed antibiotic prescription in primary care has been shown to reduce antibiotic consumption, without increasing risk of complications, yet is not widely used in the UK. We sought to quantify the relative importance of factors affecting the decision to give a delayed prescription, using a stated-choice survey among UK general practitioners. Respondents were asked whether they would provide a delayed or immediate prescription in fifteen hypothetical consultations, described by eight attributes. They were also asked if they would prefer not to prescribe antibiotics. The most important determinants of choice between immediate and delayed prescription were symptoms, duration of illness, and the presence of multiple comorbidities. Respondents were more likely to choose a delayed prescription if the patient preferred not to have antibiotics, but consultation length had little effect. When given the option, respondents chose not to prescribe antibiotics in 51% of cases, with delayed prescription chosen in 21%. Clinical features remained important. Patient preference did not affect the decision to give no antibiotics. We suggest that broader dissemination of the clinical evidence supporting use of delayed prescription for specific presentations may help increase appropriate use. Establishing patient preferences regarding antibiotics may help to overcome concerns about patient acceptance. Increasing consultation length appears unlikely to affect the use of delayed prescription.

Keywords: UK; antibiotic resistance; choice experiment; delayed prescription; general practice; primary care; respiratory tract infection; stewardship.

Figure 1

Ranking of attribute importance. Attribute descriptions: DURATION: duration of illness prior to consultation; CONSULTATION: length of consultation; PATIENT OPINION: preferences regarding antibiotics expressed by the patient; RISK FROM DELAYING: risk of harm from not starting antibiotics straight away; RISK FROM TREATMENT: risk of adverse effects from taking antibiotics; FORMAT: how the delayed prescription would be provided. Vertical axis indicates the number of respondents who ranked a given attribute at the rank shown on the horizontal axis (1 = highest rank).

Figure 2

Proportions ultimately choosing immediate, delayed and no prescription per choice question. Q: choice question, numbered by the order in which they were presented to respondents. To illustrate the patterns in proportions choosing immediate, delayed, or no prescription, the graph orders the bars by the proportion ultimately choosing the immediate prescription.

Figure 3

Coefficients for the ordered logistic regression model. Symptoms descriptions: minor throat—sore throat and swollen glands; minor chest—chesty cough and runny nose; serious throat—sore throat, swollen glands and fever; serious chest—chesty cough, fever and pain on breathing. * Reference level for categorical variables. p-values are shown where the coefficients differed (p < 0.05) between the choice to give an immediate prescription, and the choice to prescribe at all. For all other attributes and levels, the p-value for this difference was greater than 0.05, and the coefficients were constrained to be equal in the model. Bars indicate 95% confidence intervals. Abbreviations: abx—antibiotics.

Figure 4

Probability of choosing immediate, delayed or no prescription, for each symptom type. Symptom descriptions: 1 minor throat—sore throat and swollen glands; 2 minor chest—chesty cough and runny nose; 3 serious throat—sore throat, swollen glands and fever; 4 serious chest—chesty cough, fever and pain on breathing. Bars indicate 95% confidence intervals.