Recruitment and initial assessment of participants
Participants for the trial were recruited from three general practices in Nottingham familiar with research in this topic.3,4,7–9
Between September 1999 and August 2000 (excluding a month over Christmas and the millennium period), we recruited consecutive adults presenting with “acute bronchitis,” defined as a “new, acute lower respiratory tract illness in a previously well adult,” using previously reported definitions (box ).4,8–11
: Definitions for recruitment
The study was approved by the Nottingham City Hospital ethics committee, and all participants provided written consent.
Each general practitioner managed the patients according to their usual clinical practice and based on their clinical judgment divided them into two groups: group A, in which antibiotics were not definitely indicated that day, and group B, in which antibiotics were definitely indicated that day. This decision was made without additional guidance or investigations.
Antibiotic prescriptions and randomisation
All patients were given a prescription for an antibiotic, the choice of which was left to the general practitioner, and a sealed envelope containing a two week diary card with instructions, pen, and a stamped, addressed return envelope. Patients in group B were advised to take the antibiotics.
For all patients in group A the general practitioner provided verbal information based on a prompt card (box ). These patients were then randomised by using permuted blocks of four to receive or not receive a patient information leaflet about the natural course of lower respiratory tract symptoms and the advantages and disadvantages of antibiotic use (fig ). The patient information leaflet was in the sealed envelope, blinded from the general practitioner by means of a blank leaflet, together with the diary card and return envelope. Patients were asked to open and read the contents of the envelope after the consultation.
: Prompt card for verbal information given to patient by general practitioners
Information leaflet given to patients
End points and follow up
The primary end point was whether the patient took the antibiotics they had been prescribed. This information was obtained from the symptom diary, which included a space to record daily antibiotic use, and by telephone contact. Patients were contacted by telephone at around one week and two weeks after the consultation by research assistants blinded to the grouping of the patients. Answers to structured questions regarding antibiotic use were recorded.
The secondary outcome was whether patients initiated a further consultation for the same symptoms within the next month. Patients were not asked to return routinely by the general practitioner. We have previously reported that reconsultation is an easily measured and consistent end point for acute bronchitis and relates to persistent cough and patient dissatisfaction with their progress.8,9,11,12
We carried out a pilot study of 33 consecutive patients with acute bronchitis to develop consistency of data collection by the general practitioners and telephone follow up by the research assistants.
Our primary hypothesis was that the proportion of patients in group A who would take antibiotics during the two week follow up period would be lower in those who received the leaflet than in those in the control group. We calculated the risk ratio and 95% confidence interval using EpiInfo and used a χ2 test with Yates's correction for the hypothesis test. Using these data we calculated the number need to treat as the reciprocal of the absolute difference in antibiotic uptake between the two groups.
To calculate sample size we set a minimum difference of 20% in primary outcome between the two intervention arms in group A and a discriminatory power of 80%. The required number in group A was 206.
To look for possible confounding by age, sex, surgery, smoking status, description of cough, duration of cough, and the presence of chest signs we used a series of bivariate logistic regression models within Stata (version 5). We also examined whether the impact of the leaflet on antibiotic uptake was modified by any of these variables by fitting a series of multiplicative interaction terms and comparing the nested models using the likelihood ratio test.
We constructed a Kaplan-Meier plot from the days between consultation and the day antibiotics were started and calculated the rate ratio using a Cox regression model within Stata. We tested the proportional hazard assumption of this model using the diagnostic section within Stata (ph1test).