Objectives:

To assess the comparative effectiveness of interventions for improving antibiotic use for acute respiratory tract infections (RTIs) in adults and children.

Data sources:

Electronic databases (MEDLINE^® from 1990 and the Cochrane Library databases from 2005 to February 2015), reference lists of included systematic reviews, and Scientific Information Packets from point-of-care test manufacturers and experts.

Review methods:

Using predefined criteria, we selected studies of any intervention designed to improve antibiotic use for acute RTIs for which antibiotics are not indicated. Interventions were organized into education, communication, clinical, system-level, and multifaceted categories. We identified interventions that had evidence of reducing resistance to antibiotics, improving appropriate prescribing (i.e., concordant with guidelines), or decreasing overall prescribing of antibiotics for acute RTIs and not causing adverse consequences such as medical complications or patient dissatisfaction. The quality of included studies was rated and the strength of the evidence was assessed. Clinical and methodological heterogeneity limited quantitative analysis.

Results:

Although reduction in antibiotic resistance is a major goal of these interventions, there were too few studies to assess this outcome. The few studies that attempted to assess appropriate prescribing had important limitations and lack of consistency in outcome definition and ascertainment methods across studies. Therefore, reduction in overall prescribing was the only commonly reported benefit across interventions. Actual use of antibiotics was also reported in too few studies to assess separately from prescribing. No intervention had high-strength evidence for any outcome. The best evidence, from an evidence base of 133 studies, including 88 randomized controlled trials, was for four interventions with moderate-strength evidence of improved or reduced antibiotic prescribing compared with usual care that also had low-strength evidence of not causing adverse consequences. These were clinic-based parent education (21% overall prescribing reduction; similar return visits); public patient education campaigns combined with clinician education (improved appropriate prescribing; 7% reduction in overall prescribing; similar complications and satisfaction); procalcitonin for adults (12% to 72% overall prescribing reduction; similar continuing symptoms, limited activity, missing work, adverse events or lack of efficacy, treatment failure, hospitalizations, and mortality); and electronic decision support systems (improved appropriate prescribing and 5% to 9% reduction in overall prescribing; similar complications and health care use). Additionally, public parent education campaigns had low-strength evidence of reducing overall prescribing, not increasing diagnosis of complications, and decreasing subsequent visits. Other interventions had evidence of improved or reduced prescribing, but evidence on adverse consequences was lacking (streptococcal antigen testing, rapid multiviral testing in adults), insufficient (clinician and patient education plus audit and feedback plus academic detailing), or mixed (delayed prescribing, C-reactive protein [CRP] testing, clinician communication training, communication training plus CRP testing). Interventions with evidence of no impact on antibiotic prescribing were clinic-based education for parents of children 24 months or younger with acute otitis media, point-of-care testing for influenza or tympanometry in children, and clinician education combined with audit and feedback. Furthermore, limited evidence suggested that using adult procalcitonin algorithms in children is not effective and results in increased antibiotic prescribing.

Conclusions:

The best evidence supports the use of specific education interventions for patients/parents and clinicians, procalcitonin in adults, and electronic decision support to reduce overall antibiotic prescribing (and in some cases improve appropriate prescribing) for acute RTIs without causing adverse consequences, although the reduction in prescribing varied widely. Other interventions also reduced prescribing, but evidence on adverse consequences was lacking, insufficient, or mixed. Future studies should use a complex intervention framework and better evaluate measures of appropriate prescribing, adverse consequences such as hospitalization, sustainability, resource use, and the impact of potential effect modifiers.

PROSPERO number: CRD42014010094.