Eighty-three practices were randomised, and 61 of these recruited a total of 558 eligible patients between October 2006 and April 2008 (fig 1).
Fig 1 Study profile. *One patient from the control group was subsequently found to have longstanding asthma and was therefore determined (after consultation with the trial steering committee) to have been “recruited in error” and has not (more ...)
Intervention and control practices, and randomised practices that did and did not recruit participants, were similar in terms of list size, antibiotic prescribing history, and location (Wales or England) (table 1). Patients recruited by intervention and control practices were similar in terms of age, sex, duration of illness, and symptoms (table 1). Patients were recruited by intervention and control practices at a similar rate (fig 2). We achieved a follow-up rate of 94.6% (93.4% intervention, 95.8% control) for the primary outcome data. Telephone interviewers reported becoming aware of the participant’s treatment group in 34 of 509 interviews (6.7%).
Table 1 Baseline characteristics of randomised recruiting and non-recruiting practices, participating clinicians, and patients, by treatment group
Fig 2 Recruitment rates in the two groups
The number and proportion of patients experiencing each outcome, and odds ratios (with 95% confidence intervals) for the primary and secondary outcomes are shown in table 2. There was no significant difference between the intervention and control groups in the odds of reconsulting in primary care during the two weeks after registration. Children in the intervention group were significantly less likely to receive a prescription for antibiotics at the index consultation, less likely to take antibiotics during the first two weeks, and their parents were less likely to report that they would consult in the future if their child had a similar illness. There were no significant differences in terms of satisfaction, level of reassurance, parental enablement, or the parent’s rating of the “usefulness of any information received in the consultation.” Similar results were found at the univariate level, with a non-significant difference in reconsulting (absolute risk reduction 3.3%, 95% confidence interval −2.7% to 9.3%, P=0.29), and significant differences in antibiotic prescribing (absolute risk reduction 21.3%, number needed to treat 4.7, P<0.001), antibiotic consumption (absolute risk reduction 20.6%, 95% confidence interval 12.7% to 28.5%, number needed to treat 4.9, P<0.001), and future consulting intentions (21.1%, 13.1% to 29.2%, 4.7, P<0.001).
Table 2 Effect of the intervention on patient outcomes
There was no significant intervention effect when telephone consultations were counted as reconsultations along with face to face primary care consultations (odds ratio 0.81; 0.47 to 1.42), or when consultations at accident and emergency departments were included along with primary care consultations (0.85; 0.48 to 1.51). The intervention had a similar effect size on the antibiotic outcomes of receiving a prescription for antibiotics for immediate use at the index consultation (excluding prescriptions for delayed use; 0.26; 0.11 to 0.62) and receiving an antibiotic prescription at any point in the two week follow -up (0.31; 0.16 to 0.62).
The sensitivity analyses for the main reconsultation outcome and the antibiotic prescribing outcome did not result in any meaningful changes to the results (that is, there were no significant intervention effects in the sensitivity analyses for the reconsultation outcome and similar significant effects for all analyses with the antibiotic prescribing outcome).
No significant interaction effects were seen in the reconsultation models. In the antibiotic prescribing model, the intervention was more effective in above-average prescribing practices (table 3). There were no other significant interaction effects.
Table 3 Effect of practice prescribing history and study intervention on probability of being prescribed an antibiotic
Seven patients (three in the intervention group and four in the control group) were subsequently admitted to hospital or observed in a paediatric assessment unit. One patient in the control group had a longstanding diagnosis of asthma, and as such was excluded (after discussion in the trial steering committee). The longest hospital admission (two nights) was a patient in the intervention group who had febrile convulsions. The remaining admissions were one night or less.
Comparing recruited and non-recruited patients
Ninety three patients were not recruited into the study (50 were ineligible, 27 declined participation, and 16 were not recruited because of a lack of time in the consultation or for other unspecified reasons. There were no significant differences between recruited and non-recruited patients in terms of age or presenting symptoms; however, cough was more common in non-recruited patients (71.9% v 61.0%). Of the non-recruited patients, there was no significant difference in cough between the intervention and control groups. Previous duration of illness was higher in non-recruited patients than in recruited patients, although this finding was not surprising, since a duration of illness of more than seven days was a common reason for exclusion from the study.