The introduction of antibiotic therapy into medical practice has yielded substantial benefits for patients over the past 6 decades. However, the benefits of antibiotic use to individual patients come at a societal cost: the emergence of antimicrobial resistance (AMR) among bacterial pathogens. Once confined to the inpatient setting, resistant bacteria are now common community-acquired infections as well. Antibiotic use encourages the development and spread of antibiotic-resistant bacteria by at least two mechanisms: (1) by applying selective pressure, encouraging development of new strains of antibiotic-resistant bacteria, and (2) by eliminating normal bacterial flora in human hosts, which promotes colonization and spread of existing antibiotic-resistant strains. The increasing prevalence of antibiotic resistance has led to the use of more expensive and broad-spectrum antibiotics for empiric treatment of common outpatient infections, and increased morbidity and mortality among patients hospitalized with serious community-acquired infections.
Reducing inappropriate use of antibiotics is a critical step in slowing the progression of current levels of resistance, and in preventing the emergence of new strains of antibiotic-resistant bacteria. Accomplishing this requires a two-part approach. First, the use of antibiotics in conditions for which these drugs provide little or no benefit must be reduced. Second, antibiotics prescribed for patients who do require antimicrobial therapy must be appropriately targeted, and inappropriately lengthy treatment courses should be shortened.
In this fourth volume of the Closing the Quality Gap series, we critically analyze quality improvement strategies to reduce inappropriate antibiotic prescribing. We focus on interventions targeting antibiotic prescribing for acute illnesses in the outpatient setting, primarily acute respiratory infections (ARIs). Prescribing for acute conditions accounts for the majority of antibiotics dispensed in the US, and is thus likely to have the greatest influence on AMR patterns. We examined the effect of quality improvement strategies on antibiotic treatment (the decision to prescribe antibiotics for illnesses for conditions generally not requiring antibiotic therapy) and antibiotic selection (the choice of one antibiotic over another for illnesses requiring antibiotic treatment). Our review includes studies on the effect of prescribing-focused QI strategies on AMR, clinical outcomes, costs of prescribing, and patient satisfaction.
We structured the review to address the following key threshold questions:
Are quality improvement strategies to improve outpatient antibiotic use effective?
What are the critical components of effective intervention strategies?
Which patients and conditions should be targeted in order to exert the maximal impact on antibiotic prescribing?
What are the limitations of current research in this field, and which areas require further study?
As in previous reviews in this series, we performed a rigorous search of the published literature using the Cochrane Collaboration Effective Practice and Organisation of Care (EPOC) database, supplemented by targeted MEDLINE® searches. We classified QI interventions according to a modification of a taxonomy used in previous volumes of this series. The QI strategies were classified as follows:
Provision of delayed prescriptions
Audit and feedback
Clinician reminder systems
Financial or regulatory incentives for patients
Financial or regulatory incentives for clinicians
Educational strategies were subdivided into active or passive strategies, based on whether or not the learner was actively engaged in the learning process. 1 We analyzed the QI strategies for their effects on antimicrobial prescribing, defined as the change in the percentage of patient visits at which an antibiotic was prescribed (for treatment decision studies) or the change in the percentage of visits at which a recommended antibiotic was prescribed (for selection decision studies). We quantitatively synthesized the studies by determining the median effect of studies using a particular QI strategy (or combination of strategies), and rigorously evaluated for the presence of potential confounders or effect modifiers. Formal meta-analysis was not possible due to heterogeneity among the included studies. For details on the statistical methodology used, please refer to Closing the Quality Gap, Volume 1—Series Overview and Methodology (AHRQ publication No. 04-0051-1).
From a sample of 521 potentially relevant articles, we reviewed the full text of 147 articles. Of these, a total of 54 articles, reporting a total of 74 comparisons, met our inclusion criteria and were reviewed. The treatment decision was addressed in 34 articles (41 separate comparisons), and the selection decision was addressed in 26 articles (totaling 33 comparisons); six articles evaluated both the treatment and selection decision. Among these studies, 24 comparisons for the treatment decision presented data amenable to median effects analysis, and 22 comparisons for the selection decision were similarly amenable to quantitative analysis.
Based on our findings, we reached the following conclusions:
1. Quality improvement strategies are moderately effective at reducing the inappropriate prescribing of antibiotics and improving the appropriate selection of antibiotics
Overall, interventions targeting the antibiotic treatment decision were effective at reducing prescribing, with a median effect of -8.9% (interquartile range (IQR) -12.4% to -6.7%); indicating an absolute reduction in antibiotic prescribing rates of 8.9% in intervention groups compared with comparison groups. Similar effects were also apparent in studies not meeting criteria for quantitative analysis. We did not find evidence for confounding or effect modification by study design or other moderating factors.
Antimicrobial resistance was measured in only two studies, neither of which demonstrated a reduction in resistance despite a reduction in prescribing rates; however, the limited duration of followup (6 months) was likely insufficient to detect potential effects on resistance rates. Strategies to reduce antibiotic prescribing were not associated with increased use of health services, increased duration of illness symptoms, or decreased patient satisfaction. Costs were reduced in the two studies measuring this outcome.
Interventions targeting the antibiotic selection decision were also effective, with a median absolute improvement of 10.6% (IQR 3.4% to 18.2%) in prescribing of recommended antibiotics in the intervention groups compared with the comparison groups. Most studies used either clinician education alone or clinician education combined with audit and feedback. Interventions targeted both ARIs and urinary tract infections (UTIs), with some interventions targeting general prescribing; no significant differences were found for QI strategies targeting different disease processes or patient populations. In four studies, duration of antibiotic therapy for urinary tract infections was assessed, with effects ranging from no benefit to a reduction in mean antibiotic duration of approximately 2 days.
No studies assessed the effect of interventions targeting antibiotic selection on resistance, health services utilization, or disease outcomes. Three studies measuring prescribing costs showed 20–30% relative reductions attributable to the intervention.
We conclude that QI strategies are moderately effective at improving prescribing behavior, with regard to both antibiotic treatment and antibiotic selection decisions. Organizations should compare studies performed in similar settings, and with similar patient and provider populations, in order to identify specific QI strategies that are most likely to be effective in their own setting. Several particularly salient studies (representative of the types of interventions and settings encountered in our review) are summarized in Appendix A * , to provide a starting point for considering important implementation factors. Appendix B contains summaries of all included studies organized by setting and patient population in order to facilitate identification of relevant studies.
2. Although no single quality improvement strategy is clearly superior, active clinician education may be more effective in certain settings
Studies predominantly used clinician education strategies or clinician education in combination with patient education; smaller numbers of studies used audit and feedback, or other combinations of strategies. We did not find definitive evidence for superiority of one strategy (or combination of strategies) over another, within either the antibiotic treatment studies or antibiotic selection studies. Delayed prescribing interventions achieved large absolute reductions in ultimate antibiotic consumption, but these studies operated under the assumption that the “default” practice was to uniformly administer antibiotics (even when they might not be indicated, as in acute cough illness). In the treatment studies, the trend suggested that active educational strategies were more effective (p=0.11), an effect also seen in studies not eligible for median effects analysis. Among antibiotic selection studies, the addition of audit and feedback conferred significantly less benefit than clinician education alone (3.4% vs. 13.9%, P=0.03). This finding may in part be explained by confounding: clinician education-alone studies were more likely to have a small sample size, and smaller studies were associated with larger median effects. This effect may be mediated through publication bias, a higher level of engagement between the study directors and target physicians and/or greater intensity of the intervention. In contrast to interventions aimed at the treatment decision, interventions using active educational strategies to improve antibiotic selection were not significantly associated with larger median effects. However, in each of the five studies where active and passive educational strategies were compared head-to-head, the active strategies were superior.
3. Interventions targeting prescribing for all acute respiratory tract infections may exert a greater effect on overall prescribing than interventions targeting specific types of acute respiratory infections
Most treatment studies targeted prescribing for ARIs; there were no significant differences in effectiveness for interventions targeting specific ARIs vs. general ARIs, or for interventions targeting children vs. adults. Among antibiotic selection studies, there was no difference in median effects among interventions targeting general ARIs, specific ARIs, or UTIs. In an effort to maximize the relevance of interventions targeting the antibiotic treatment decision at the health system level, we conducted a separate analysis extrapolating study effect sizes to the population level, determining the number of antibiotics per 1000 person-years that could be saved as a result of the intervention. Interventions targeting prescribing for all ARIs have more impressive effects when their impact is extrapolated to the population level (35–85 antibiotic prescriptions saved per 1000 person-years) even though the individual study effect sizes are more modest. In contrast, condition-specific interventions (e.g., those targeting prescribing for pharyngitis in children), reporting large effects within the study population, have smaller effects on prescribing at the population level (e.g., 5–15 antibiotics saved per 1000 person-years). The greater population effects of strategies targeting prescribing for all ARIs (as opposed to focusing on a single condition or patient age range) could result in relatively higher cost savings as well, depending on the nature and intensity of the intervention.
4. Study design and quality should be improved. Studies that formally evaluate the cost effectiveness of interventions to improve antibiotic treatment and selection are needed, and studies should evaluate the potential harms of such interventions
As noted in previous reviews of diabetes and hypertension QI strategies in this series, 1, 2 the methodological quality of included trials was only fair. In both treatment and selection studies, approximately half the trials were non-randomized; most failed to document the rationale for selection of the comparison group. Other basic quality problems such as inadequate concealment of allocation and unit-of-analysis errors appeared frequently. Moreover, studies consistently failed to describe the theoretical basis for their intervention or the rationale for choice of the QI strategies used, and did not document the reach of the intervention (i.e., the extent to which the target population actually received the intervention). These differences may affect the internal and external validity of the studies. In addition, it is also likely that our results were confounded by substantial inter-study variations in patient population, disease target, and clinical settings, as well as many unmeasured confounding variables such as local factors, the culture of medical practice, and the health care system structure.
Future studies should aim for higher methodological standards, and should maximize the applicability of their interventions by clearly defining the characteristics of the intervention, the setting, and the participants. Few studies evaluated potential harms to patients that could result from reducing antibiotic use, such as adverse clinical consequences or increased use of health services. In addition, our conclusions are limited to fairly short term intervention effects. Most importantly, very few studies reported any information on the implementation costs, and no studies performed a formal cost effectiveness analysis. Documentation that QI strategies recover their implementation costs through savings in antibiotic costs would greatly enhance the appeal of such programs to health systems.
Active clinician education strategies included academic detailing (educational outreach), consensus-building sessions, and educational workshops. Active patient education strategies included one-on-one or group educational meetings. Passive educational strategies (for both clinicians and patients) included distribution of educational materials (e.g., waiting room pamphlets for patients) and lectures (e.g., traditional CME for clinicians).
* Appendixes cited in this report are provided electronically at http://www
.ahrq .gov/downloads/pub /evidence/pdf/medigap/medigap.pdf
Agency for Healthcare Research and Quality (US), Rockville (MD)
Ranji SR, Steinman MA, Shojania KG, et al. Closing the Quality Gap: A Critical Analysis of Quality Improvement Strategies (Vol. 4: Antibiotic Prescribing Behavior). Rockville (MD): Agency for Healthcare Research and Quality (US); 2006 Jan. (Technical Reviews, No. 9.4.) Executive Summary.