Antimicrobial overuse, resistance to existing drugs, and the paucity of new agents under development have combined to form what the Centers for Disease Control and Prevention has termed “one of our most serious health threats.” 1 The majority of antimicrobials administered to humans are prescribed in outpatient settings, and overuse is common. Approximately 80% of adults with rhinosinusitis are prescribed antimicrobials,Reference Fairlie, Shapiro, Hersh and Hicks 2 , Reference Smith, Kern, Chandra, Tan and Evans 3 and >60% of patients with pharyngitis received antimicrobials despite data suggesting that only 10% have an antimicrobial-responsive infection.Reference Barnett and Linder 4 Factors contributing to high rates of prescribing include patient expectations, patient and provider unawareness of antimicrobial resistance, and lack of appreciation regarding the seriousness of the threat posed by antimicrobial resistance.Reference Costelloe, Metcalfe, Lovering, Mant and Hay 5
Antimicrobial stewardship programs (ASPs) are focused efforts by a health care system or a part of the system (eg, an outpatient clinic) to optimize antimicrobial use. Goals of ASPs include improving patient outcomes, decreasing negative consequences including adverse drug reactions and antimicrobial-associated infections (eg, Clostridium difficile infection), limiting antimicrobial resistance, and delivering cost-effective therapy.Reference MacDougall and Polk 6 – Reference Ohl and Dodds Ashley 9
In a previous review,Reference Ranji, Steinman and Shojania 10 – Reference Steinman, Ranji, Shojania and Gonzales 12 quality improvement strategies (primarily clinician and/or patient education) were found to be moderately effective in reducing inappropriate antimicrobial prescribing and improving appropriate antimicrobial selection, but few studies reported patient or microbial outcomes. We conducted a systematic review of the recent evidence regarding the effectiveness of ASPs in outpatient settings, with an emphasis on patient outcomes and microbial outcomes, and including the more commonly reported prescribing outcomes. To avoid overlap with the existing review, we excluded any studies cited in the full Technical ReviewReference Ranji, Steinman and Shojania 10 or related publications.Reference Ranji, Steinman, Shojania and Gonzales 11 , Reference Steinman, Ranji, Shojania and Gonzales 12 This report is derived from work performed for a larger Department of Veterans Affairs Evidence-based Synthesis Program review.
Methods
Search Strategy
We based our search strategy on Cochrane reviews of antimicrobial stewardshipReference Davey, Brown and Fenelon 13 , Reference Arnold and Straus 14 and searched MEDLINE (Ovid) from 2000 through November 2013, limited to English language studies enrolling human subjects (Appendix). We identified additional studies from the Cochrane Library, systematic reviews, reference lists, and suggestions from peer reviewers of the Evidence-based Synthesis Program review.
Study Selection
Titles, abstracts, and articles were reviewed by investigators and research associates. Included studies were (1) conducted in settings or enrolling patients relevant to the United States (eg, patients with infections likely in the United States; settings where antimicrobials are available only by prescription); (2) involving an intervention of interest with an assessment of intervention effects; (3) reporting outcomes of interest; (4) not involving prophylactic antimicrobials; (5) involving patients with bacterial (vs viral, fungal, or mycobacterial) infections; and (6) randomized controlled trials (RCTs), cluster randomized controlled trials (CRCTs), controlled clinical trial (CCTs), controlled before/after trial (CBAs), or interrupted times series (ITS) with at least 3 data points before and after intervention implementation. Interventions which did not meet inclusion criteria include national campaigns to educate clinicians and patients regarding optimizing antimicrobial use. These interventions are not implemented at the institution or system level, and thus were considered beyond the scope of this review.
Data Extraction and Synthesis
From eligible studies, we extracted study characteristics, outcomes (prescribing, patient, and microbial), costs, and harms. Categorization measures considered the primary focus of the intervention as described by study authors. Prescribing outcomes included percentage of subjects receiving antimicrobials, drug selection, therapy duration, and guideline-concordant use. Patient outcomes included return visits, hospitalizations, adverse events, delayed antimicrobial prescriptions, and patient satisfaction. Information regarding barriers to implementation, sustainability, and scalability was recorded. Data extraction was verified by the lead author. For categorical data, we report odds and risk ratios. For continuous data we report mean or median differences. From ITS studies, we report, where provided by study authors, level and trend (or slope) results.
We assessed risk of bias for individual studies using criteria developed for the Cochrane Effective Practice and Organization of Care reviews. 15 A study was rated as low risk if each of the individual criteria were scored as low, medium risk if one or two criteria were scored as unclear or high, and high risk if more than two criteria were scored as unclear or high. Quality of an existing systematic review was assessed using the measurement tool for assessment of multiple systematic reviews.Reference Shea, Grimshaw and Wells 16
We rated overall strength of evidence (high, medium, low, or insufficient) for prescribing, patient, and microbial outcomes for each intervention category using methods developed by the Agency for Healthcare Research & Quality (AHRQ) and the Effective Health Care Program.Reference Owens, Lohr and Atkins 17 Strength of evidence was evaluated based on four domains: (1) risk of bias, (2) consistency, (3) directness, and (4) precision. Due to heterogeneity of interventions, study designs, patient populations, and outcomes reporting, results could not be accurately pooled. We compiled a summary of findings and drew conclusions based on qualitative synthesis of the findings. To minimize publication bias, we performed a comprehensive literature search, hand searched reference lists, and received input from content experts; however, funnel plots were not possible due to the small number of trials for each intervention.
Results
We reviewed 6,694 titles and abstracts from the literature search. We excluded 6,125 after abstract review and an additional 529 after full text review, leaving 40 articles eligible for inclusion (Figure). Hand searching or reviewer suggestion identified 10 further articles, totaling 50 included articles (17 RCTs, 18 CRCTs, 3 CCTs, 6 CBA trials, and 6 ITS studies).Reference Gerber, Prasad and Fiks 18 – Reference Takemura, Ebisawa and Kakoi 69 Studies were conducted in the United States or Canada (N=21),Reference Gerber, Prasad and Fiks 18 – Reference Worrall, Hutchinson, Sherman and Griffiths 37 Europe or the United Kingdom (N=24),Reference Butler, Simpson and Dunstan 38 – Reference Cals, Schot, de Jong, Dinant and Hopstaken 64 the Middle East (N=3),Reference Regev-Yochay, Raz and Dagan 65 – Reference Chazan, Turjeman and Frost 67 and the Asia/Pacific region (N=2).Reference Pagaiya and Garner 68 , Reference Takemura, Ebisawa and Kakoi 69 Of these, 14 studies that enrolled adults,Reference Metlay, Camargo and MacKenzie 21 , Reference Gonzales, Sauaia and Corbett 22 , Reference Worrall, Kettle, Graham and Hutchinson 27 , Reference Manns, Laupland, Tonelli, Gao and Hemmelgam 30 , Reference Marshall, Gough and Grootendorst 31 , Reference Rattinger, Mullins and Zukerman 35 , Reference Worrall, Hutchinson, Sherman and Griffiths 37 , Reference Slekovec, Leroy and Vernaz-Hegi 48 – Reference Seager, Howell-Jones, Dunstan, Lewis, Richmond and Thomas 50 , Reference Little, Moore and Turner 52 , Reference Cals, deBock and Beckers 56 5 enrolled children or adolescents,Reference Gerber, Prasad and Fiks 18 , Reference Finkelstein, Huang and Kleinman 20 , Reference Francis, Butler, Hood, Simpson, Wood and Nuttall 57 , Reference Regev-Yochay, Raz and Dagan 65 , Reference Pagaiya and Garner 68 and 31 enrolled all ages or did not specify age. Most enrolled patients with respiratory infections (29 trials). Summary data on prescribing and patient outcomes are presented in Tables 1 and 2; no study reported microbial outcomes. Although study heterogeneity precluded pooling results, the effects of individual studies are presented in Supplemental Tables 1 and 2, along with strength of evidence.
Figure Literature Flow Diagram
Table 1 Overview of Prescribing Outcomes—Antimicrobial Stewardship Interventions for Outpatients
NOTE. ASP, antimicrobial stewardship program; NR, not reported; PCR, polymerase chain reaction; CRP, C-reactive protein; CBA, controlled before and after; CCT, controlled clinical trial; CRCT, cluster randomized controlled trial; ITS, interrupted time series; RCT, randomized controlled trial.
+Indicates statistically significant difference favoring antimicrobial stewardship intervention.
≈Indicates no statistically significant difference between antimicrobial stewardship intervention and control.
−Indicates statistically significant difference favoring control.
+/−Indicates mixed results across different antimicrobials studied or differences between level and trend outcomes in ITS analyses.
a Some studies with a “+” reported mixed results (ie, significant differences for some conditions or some age groups, no difference for others).
b Includes 1 study with significance not reported.
c Decreased antimicrobial use was also reported in 2 studies from an existing systematic review.
Table 2 Overview of Patient Outcomes—Antimicrobial Stewardship Interventions for Outpatients
NOTE. ASP, antimicrobial stewardship program; NR, not reported; PCR, polymerase chain reaction; CRP, C-reactive protein; CBA, controlled before and after; CCT, controlled clinical trial; ITS, interrupted time series; RCT, randomized controlled trial.
+Indicates statistically significant difference favoring antimicrobial stewardship intervention.
≈Indicates no statistically significant difference between antimicrobial stewardship intervention and control.
−Indicates statistically significant difference favoring control.
Effectiveness of interventions on prescribing and patient outcomes
Provider and/or Patient Education
In 16 studies of provider and/or patient education (5 RCTs,Reference Butler, Simpson and Dunstan 38 , Reference Varonen, Rautakorpi and Nyberg 43 , Reference Little, Rumsby and Kelly 44 , Reference Chazan, Turjeman and Frost 67 , Reference Pagaiya and Garner 68 6 CRCTs,Reference Gerber, Prasad and Fiks 18 , Reference Finkelstein, Huang and Kleinman 20 , Reference Metlay, Camargo and MacKenzie 21 , Reference van Driel, Coenen and Dirven 42 , Reference Regev-Yochay, Raz and Dagan 65 , Reference Esmaily, Silver and Shiva 66 1 CCT,Reference Gonzales, Sauaia and Corbett 22 and 4 CBAsReference Vinnard, Linkin and Localio 19 , Reference Stewart, Pilla and Lunn 23 , Reference Légaré, Labrecque and LeBlanc 29 , Reference Llor, Bjerrum and Arranz 40 , Reference Smeets, Kuyvenhoven and Akkerman 41 ), interventions were directed at providers in 13 of 16 studies and ranged from single to multiple sessions. Most provider education interventions were multifaceted and included discussion of current guidelines, feedback, patient education, communications skills training, or information regarding C-reactive protein (CRP) testing.
Antimicrobial prescribing was reported in 15 studies.Reference Gerber, Prasad and Fiks 18 – Reference Stewart, Pilla and Lunn 23 , Reference Butler, Simpson and Dunstan 38 – Reference van Driel, Coenen and Dirven 42 , Reference Little, Rumsby and Kelly 44 , Reference Regev-Yochay, Raz and Dagan 65 – Reference Pagaiya and Garner 68 Of these, 6 found decreased prescribingReference Gerber, Prasad and Fiks 18 , Reference Finkelstein, Huang and Kleinman 20 , Reference Metlay, Camargo and MacKenzie 21 , Reference Butler, Simpson and Dunstan 38 , Reference Regev-Yochay, Raz and Dagan 65 , Reference Chazan, Turjeman and Frost 67 and 6 found no difference.Reference Vinnard, Linkin and Localio 19 , Reference Gonzales, Sauaia and Corbett 22 , Reference Smeets, Kuyvenhoven and Akkerman 41 , Reference van Driel, Coenen and Dirven 42 , Reference Little, Rumsby and Kelly 44 , Reference Esmaily, Silver and Shiva 66 Of the remaining 3 studies, 1 study reported decreased prescribing for lower respiratory tract infections but not acute rhinosinusitis,Reference Llor, Cots and López-Valcárcel 39 , Reference Llor, Bjerrum and Arranz 40 1 study reported decreased prescribing for respiratory infections but not diarrhea,Reference Pagaiya and Garner 68 and 1 study reported a 9.4% decrease in total antimicrobial prescribing during the study, but the significance of this finding was not reported.Reference Stewart, Pilla and Lunn 23
Patient outcomes were reported in 3 studies. In 1 RCT, a higher number of return clinic visits per patient was observed during the month after the initial visit in the group receiving a patient education leaflet.Reference Little, Rumsby and Kelly 44 , Reference Moore, Little, Rumsby and Kelly 70 No differences in hospitalizations (2 studies),Reference Metlay, Camargo and MacKenzie 21 , Reference Butler, Simpson and Dunstan 38 adverse events (1 study),Reference Little, Rumsby and Kelly 44 or satisfaction with care (1 study)Reference Metlay, Camargo and MacKenzie 21 were observed.
Provider Feedback
In 3 of the 5 studies of provider feedback (1 RCT,Reference Naughton, Feely and Bennett 46 2 CRCTs,Reference Linder, Schnipper and Tsurikova 24 , Reference Gjelstad, Høye, Straand, Brekke, Dalen and Lindbæk 45 1 CCT,Reference Madridejos-Mora, Amado-Guirado and Pérez-Rodriguez 47 and 1 CBAReference Vinnard, Linkin and Localio 19 ), individualized feedback regarding antimicrobial prescribing was associated with significant decreases in prescribing compared to more general feedback or usual care.Reference Vinnard, Linkin and Localio 19 , Reference Gjelstad, Høye, Straand, Brekke, Dalen and Lindbæk 45 , Reference Madridejos-Mora, Amado-Guirado and Pérez-Rodriguez 47 Prescribing outcomes were similar when postal feedback plus academic detailing (outreach visit from the research coordinator) was compared to postal feedback alone,Reference Naughton, Feely and Bennett 46 or when an electronic record component was compared to usual care.Reference Linder, Schnipper and Tsurikova 24 No study reported patient outcomes.
Guidelines
Antimicrobial prescribing guidelines were assessed in 6 studies (1 CRCT,Reference Seager, Howell-Jones, Dunstan, Lewis, Richmond and Thomas 50 1 CCT,Reference Martens, Winkens, van der Weijden, de Bruyn and Severens 51 4 ITSReference Dowell, Tian and Stover 25 , Reference Weiss, Blais, Fortin, Lantin and Gaudet 26 , Reference Slekovec, Leroy and Vernaz-Hegi 48 , Reference Venekamp, Rovers, Verheij, Bonten and Sachs 49 ) for urinary tract infections (UTIs),Reference Slekovec, Leroy and Vernaz-Hegi 48 sexually transmitted infections,Reference Dowell, Tian and Stover 25 acute dental pain,Reference Seager, Howell-Jones, Dunstan, Lewis, Richmond and Thomas 50 acute rhinosinusitis,Reference Venekamp, Rovers, Verheij, Bonten and Sachs 49 and overall antimicrobial use.Reference Weiss, Blais, Fortin, Lantin and Gaudet 26 , Reference Martens, Winkens, van der Weijden, de Bruyn and Severens 51 In 4 studies detailing antimicrobial use following guideline introduction, 3 found significant decreases post-intervention.Reference Weiss, Blais, Fortin, Lantin and Gaudet 26 , Reference Venekamp, Rovers, Verheij, Bonten and Sachs 49 , Reference Seager, Howell-Jones, Dunstan, Lewis, Richmond and Thomas 50 In 1 study of guidelines to improve antimicrobial selection, mixed results across antimicrobials were reported,Reference Slekovec, Leroy and Vernaz-Hegi 48 while another reported no difference in patient satisfaction between those who did or did not receive an antimicrobial.Reference Seager, Howell-Jones, Dunstan, Lewis, Richmond and Thomas 50
Delayed Prescribing
Delayed prescribing was assessed in 4 RCTs,Reference Worrall, Kettle, Graham and Hutchinson 27 , Reference Little, Rumsby and Kelly 44 , Reference Little, Moore and Turner 52 , Reference Cals, Schot, de Jong, Dinant and Hopstaken 64 wherein providers ask patients to fill a prescription only if symptoms persist or worsen. In 2 studies, delayed prescribing was the primary intervention. A significant reduction in antimicrobial use was found in 1 study of women with UTIs who received delayed prescriptions versus those who received immediate prescriptions.Reference Little, Moore and Turner 52 A second study found no significant difference in prescriptions filled when patients were given post-dated (2-d delay) versus same-day prescription.Reference Worrall, Kettle, Graham and Hutchinson 27
Two other studies included a delayed prescribing component. One study, summarized under Provider and/or Patient Education because it included education versus no education groups, reported a significant reduction in antimicrobial use in the group assigned to delayed prescribing versus the immediate antimicrobial group.Reference Little, Rumsby and Kelly 44 Another study, summarized under Laboratory Tests (below) because it included CRP testing, found fewer patients in the CRP group receiving delayed prescriptions filled the prescriptions, versus control patients, who also received delayed prescriptions.Reference Cals, Schot, de Jong, Dinant and Hopstaken 64
One study found lower odds of return clinic visits in the delayed prescription group compared with immediate prescription for women with urinary tract infections (UTIs);Reference Little, Moore and Turner 52 there were no major adverse events in either group. Another found that return clinic visits did not differ between groups assigned to delayed or immediate antimicrobial prescriptions.Reference Little, Rumsby and Kelly 44
Communication Skills Training
Communication skill training for providers is intended to enhance patient–provider communication, address patient expectations for antimicrobial treatment, and foster a more “patient-centered” approach to care. All of the included studies (6 CRCTsReference Légaré, Labrecque, Cauchon, Castel, Turcotte and Grimshaw 28 , Reference Légaré, Labrecque and LeBlanc 29 , Reference Little, Stuart and Francis 53 – Reference Altiner, Brockmann, Sielk, Wilm, Wegscheider and Abholz 58 ) involved multifaceted interventions. Of 6 eligible studies, 5 studies reported significantly reduced antimicrobial prescribing following the intervention.Reference Légaré, Labrecque, Cauchon, Castel, Turcotte and Grimshaw 28 , Reference Little, Stuart and Francis 53 – Reference Altiner, Brockmann, Sielk, Wilm, Wegscheider and Abholz 58 For patient outcomes, the return clinic visit rate did not differ between intervention and control (3 studies).Reference Légaré, Labrecque, Cauchon, Castel, Turcotte and Grimshaw 28 , Reference Cals, Butler, Hopstaken, Hood and Dinant 54 – Reference Francis, Butler, Hood, Simpson, Wood and Nuttall 57 Patient satisfaction was mixed, with improved satisfaction in the intervention group in 1 of 4 studies.Reference Légaré, Labrecque, Cauchon, Castel, Turcotte and Grimshaw 28
Restriction Policies
Two studies (2 ITSReference Manns, Laupland, Tonelli, Gao and Hemmelgam 30 , Reference Marshall, Gough and Grootendorst 31 ) assessed restriction policies. One was of fluoroquinolone restriction,Reference Manns, Laupland, Tonelli, Gao and Hemmelgam 30 which was not associated with any significant change in the rate of fluoroquinolone prescribing but was associated with a significant increase in prescriptions consistent with formulary guidelines. There were no changes in mortality or infection-related hospitalizations, but small statistically significant increases in both return clinic visits and all-cause hospitalization were observed. A second study evaluated the effects of limiting reimbursement for fluoroquinolones to treatment of patients with specified conditions.Reference Marshall, Gough and Grootendorst 31 In this study, a decreasing trend in total antimicrobial prescriptions followed the introduction of the restriction policy, with mixed results for specific antimicrobials.
Computerized Clinical Decision Support
Computerized clinical decision support within an electronic medical record was evaluated in 6 studies (2 RCTs,Reference Jenkins, Irwin and Coombs 33 , Reference McGinn, McCullagh and Kannry 34 3 CRCTs,Reference Gonzales, Anderer and McCulloch 32 , Reference Linder, Schnipper and Tsurikova 36 , Reference Martens, van der Weijden and Severens 59 1 CBAReference Rattinger, Mullins and Zukerman 35 ), and was associated with decreased prescribing in 4 of 6.Reference Gonzales, Anderer and McCulloch 32 – Reference Rattinger, Mullins and Zukerman 35 Of the 2 remaining studies, 1 study reported no difference but also reported low uptake of the decision support by providers,Reference Linder, Schnipper and Tsurikova 36 while another reported mixed results. Reminders were associated with increased adherence to only some of the prescribing recommendations.Reference Martens, van der Weijden and Severens 59 Among patient outcomes, no significant differences were reported for return clinic visits (4 studies),Reference Gonzales, Anderer and McCulloch 32 – Reference McGinn, McCullagh and Kannry 34 , Reference Linder, Schnipper and Tsurikova 36 hospitalization (2 studies),Reference Gonzales, Anderer and McCulloch 32 , Reference Jenkins, Irwin and Coombs 33 delayed antimicrobial prescriptions (2 studies),Reference Jenkins, Irwin and Coombs 33 , Reference McGinn, McCullagh and Kannry 34 or adverse events (1 study).Reference Gonzales, Anderer and McCulloch 32
Financial Incentives
A single study (CBA) described a one-time payment (independent of practice performance) improving the volume of prescribing and adherence to guidelines for just 2 of 7 antimicrobials studied; these researchers also noted that changes diminished during the first year.Reference Martens, Werkhoven and Severens 60
Procalcitonin, Rapid Antigen Detection Tests, Polymerase Chain Reaction Assay, and C-Reactive Protein
A high-quality systematic review found that procalcitonin testing in patients with acute respiratory tract infection was associated with decreased antimicrobial prescriptions.Reference Schuetz, Miller, Christ-Crain, Stoltz, Tamm and Bouadma 71 In more recent studies (6 RCTs,Reference Worrall, Hutchinson, Sherman and Griffiths 37 , Reference Little, Hobbs and Moore 61 – Reference Cals, Schot, de Jong, Dinant and Hopstaken 64 , Reference Takemura, Ebisawa and Kakoi 69 2 CRCTs,Reference Little, Stuart and Francis 53 – Reference Cals, deBock and Beckers 56 and 1 CBAReference Llor, Cots and López-Valcárcel 39 , Reference Llor, Bjerrum and Arranz 40 ), rapid antigen detection and viral polymerase chain reaction (PCR) testing in patients with acute respiratory tract infection were associated with an initial decrease in antimicrobial prescriptions, although this was not sustained throughout the study period.Reference Brittain-Long, Westin, Olofsson, Lindh and Andersson 62 Testing for Group A Streptococcus antigen, either alone or in combination with pharyngitis decision rules, was associated with decreased antimicrobial prescriptions compared to usual care.Reference Worrall, Hutchinson, Sherman and Griffiths 37 A second study of rapid antigen testing for patients with pharyngitis found that rapid testing combined with a clinical score was associated with decreased antimicrobial use compared to delayed prescribing, but the rapid test did not provide additive value to the clinical score alone.Reference Little, Hobbs and Moore 61
Of 6 studies of CRP testing (alone and in combination with communication skills training) in patients with acute respiratory tract or mixed infections, 5 studies showed decreased antimicrobial prescriptions and avoidance of newer broad-spectrum antimicrobials in select patients.Reference Llor, Cots and López-Valcárcel 39 , Reference Llor, Bjerrum and Arranz 40 , Reference Little, Stuart and Francis 53 , Reference Cals, Butler, Hopstaken, Hood and Dinant 54 – Reference Cals, deBock and Beckers 56 , Reference Cals, Schot, de Jong, Dinant and Hopstaken 64 , Reference Takemura, Ebisawa and Kakoi 69
No differences were observed between groups receiving any of the tests studied and comparator groups in return clinic visits,Reference Cals, Butler, Hopstaken, Hood and Dinant 54 – Reference Cals, deBock and Beckers 56 , Reference Little, Hobbs and Moore 61 , Reference Diederichsen, Skamling and Diederichsen 63 , Reference Cals, Schot, de Jong, Dinant and Hopstaken 64 , Reference Takemura, Ebisawa and Kakoi 69 hospitalizations,Reference Cals, Butler, Hopstaken, Hood and Dinant 54 – Reference Cals, deBock and Beckers 56 , Reference Little, Hobbs and Moore 61 , Reference Cals, Schot, de Jong, Dinant and Hopstaken 64 , Reference Takemura, Ebisawa and Kakoi 69 modification of initial treatment,Reference Takemura, Ebisawa and Kakoi 69 duration of fever,Reference Takemura, Ebisawa and Kakoi 69 or performance of further testing.Reference Takemura, Ebisawa and Kakoi 69 CRP testing and communication skills training were associated with similar, or possibly increased, patient satisfaction with care.Reference Cals, Butler, Hopstaken, Hood and Dinant 54 – Reference Cals, deBock and Beckers 56 , Reference Cals, Schot, de Jong, Dinant and Hopstaken 64
Costs
Dispensing costs were reported in 7 studies (3 RCTs,Reference Butler, Simpson and Dunstan 38 , Reference Chazan, Turjeman and Frost 67 , Reference Pagaiya and Garner 68 1 CCT,Reference Madridejos-Mora, Amado-Guirado and Pérez-Rodriguez 47 1 CRCT,Reference Cals, Butler, Hopstaken, Hood and Dinant 54 , Reference Cals, Ament and Hood 55 and 2 ITSReference Weiss, Blais, Fortin, Lantin and Gaudet 26 , Reference Marshall, Gough and Grootendorst 31 ). Significant cost reductions associated with ASPs were found in 1 study of provider education,Reference Pagaiya and Garner 68 a study of provider feedback,Reference Madridejos-Mora, Amado-Guirado and Pérez-Rodriguez 47 and a study of guidelines for common infectious conditions.Reference Weiss, Blais, Fortin, Lantin and Gaudet 26 One study of provider education reported greater savings with continuous medical education compared with seasonal medical education.Reference Chazan, Turjeman and Frost 67 A “limited use” policy was associated with mixed findings (ie, decreased costs for some antimicrobials but not others).Reference Marshall, Gough and Grootendorst 31 One study reported that medication cost per patients decreased with communication skills training and with CRP testing.Reference Cals, Butler, Hopstaken, Hood and Dinant 54 , Reference Cals, Ament and Hood 55
Three studies reported program costs (2 RCTsReference Butler, Simpson and Dunstan 38 , Reference Naughton, Feely and Bennett 46 and 1 CRCTReference Cals, Butler, Hopstaken, Hood and Dinant 54 , Reference Cals, Ament and Hood 55 ). A study of a provider education program reported a mean cost per practice of £2,923 (US$4,860 in 2014) covering administration costs, seminar preparation and seminar delivery.Reference Butler, Simpson and Dunstan 38 A study of provider feedback reported total cost per practice of €175 (US$243) covering staff, equipment, and administrative costs.Reference Naughton, Feely and Bennett 46 The study of communication skills training and CRP testing reported per patient program costs ranging from €0.00 (usual care) to €10.06 (US$13.95) (combined CRP plus communication skills training).Reference Cals, Butler, Hopstaken, Hood and Dinant 54 , Reference Cals, Ament and Hood 55
Key Intervention Components
Information on key intervention components is limited. Speculation by authors or reported data from individual providerReference Finkelstein, Huang and Kleinman 20 , Reference Metlay, Camargo and MacKenzie 21 , Reference Aagaard, Gonzales and Camargo 72 , Reference Stille, Rifas-Shiman, Kleinman, Kotch and Finkelstein 73 or focus groupReference Gjelstad, Høye, Straand, Brekke, Dalen and Lindbæk 45 , Reference Frich, Høye, Lindbæk and Straand 74 interviews suggested that leadership, a team approach, patient education materials, provider reminders, user-friendly interfaces, and evidence-based materials may be key, but little evidence was presented to support such claims.
Effectiveness by Clinic Setting or Suspected Patient Condition
All but 7 studies were conducted in primary care clinics. Respiratory infections were the condition-of-interest in 29 studies. We found little information regarding the effectiveness of stewardship interventions in other settings or infections.
Harms of Antimicrobial Stewardship Programs in Outpatient Settings
No studies were powered to detect between-group differences in harms. In total, 20 studies reported return clinic visits, hospitalizations, and/or adverse events including mortality and only 3 found significant differences between intervention and control groups.Reference Manns, Laupland, Tonelli, Gao and Hemmelgam 30 , Reference Little, Rumsby and Kelly 44 , Reference Little, Moore and Turner 52
Implementation Facilitators
In several studies, stewardship implementation was addressed. Providers reported being more likely to utilize a computer-based intervention if it was easy to access, similar to existing software, and not overly complex.Reference Linder, Schnipper and Tsurikova 36 Similarly, convenient location and scheduling, interactive sessions, evidence-based information, and relevant topics were mentioned by participants as being important.Reference Allaire, Labrecque, Giguere, Gagnon and Légaré 75
Strength of Evidence
Only the associations between prescribing outcomes and communication skills training and laboratory testing were supported by medium-strength evidence (Table 3). Strength of evidence was low for associations between prescribing outcomes and other interventions, and it was low or insufficient for patient outcomes. Details regarding strength of evidence are presented in Supplemental Tables 1 (prescribing outcomes) and 2 (patient outcomes).
Table 3 Overview of Strength of Evidence—Antimicrobial Stewardship Interventions for Outpatients
NOTE. ASP, antimicrobial stewardship program.
a Number of studies is >50; studies with multiple interventions are included under each intervention.
Discussion
Our systematic review provides updated information on the impact of outpatient ASPs on prescribing, patient, microbial, and cost outcomes. We identified several main findings. First, outpatient antimicrobial stewardship interventions of all types were associated with favorable changes in antimicrobial prescribing. Second, changes in prescribing did not adversely affect patient outcomes or drug costs, although these outcomes were not universally reported. Third, no study reported the effect of outpatient stewardship interventions on microbial outcomes. Fourth, studies of outpatient antimicrobial stewardship predominantly involve respiratory infections; therefore, we have little information is available regarding antimicrobial stewardship for other common outpatient infections. Importantly, few interventions were supported by medium-strength evidence, and none by high-strength evidence. Additionally, many interventions in the included studies were multifaceted, and few provided separate results for different intervention components.
Given the high rate of unnecessary prescribing for respiratory infections in outpatient settings, it is not surprising that the majority of included studies were designed to address that concern. As a result, little information is available regarding whether the stewardship interventions would be effective with other infections or settings, including common infections such as cellulitis or other skin/soft-tissue infections. Urinary tract infections, which are commonly misdiagnosed and overtreated, are also underrepresented in studies of antimicrobial stewardship. These conditions may represent promising areas for both achieving further reductions in antimicrobial use and further stewardship studies. In addition to the lack of data on nonrespiratory infections, we also found limited information on scalability and sustainability of interventions. While many interventions were conducted at multiple sites, few were replicated or provided long-term results after the initial research team was no longer present. Future research should focus on assessing long-term sustainable improvements in clinically meaningful outcomes, should expand stewardship programs to non-respiratory infections, and should assess patient and microbial outcomes, in addition to the usual prescribing outcomes.
Laboratory testing to aid antimicrobial stewardship, especially procalcitonin and CRP assays, appears to be a promising tool that can be used to significantly decrease antimicrobial prescribing. Their objective results are likely a welcome aid to clinicians, who often must assess the risk/benefit ratio of antimicrobial treatment largely on subjective data. Similarly, efforts to improve provider communication with patients around antimicrobial use also showed promising results. As efforts from the CDC and others to educate the public about the growing risk of antimicrobial resistance 1 reach more patients, clinicians may find that patients will respond even more favorably to communication-based stewardship efforts.
In conclusion, a wide variety of stewardship efforts are associated with decreased antimicrobial prescribing, without evidence of harms or increased costs, although these outcomes were not universally assessed. Importantly, in this era of increasing antimicrobial resistance, the effects of outpatient stewardship programs on antimicrobial resistance are unknown. However, the ecological evidence linking increasing antimicrobial use and antimicrobial resistance is robust and biologically plausible.Reference van de Sande-Bruinsma and Grundmann 76 Accordingly, the reductions in antimicrobial use are likely to offer a clinical benefit—albeit one that is yet unquantified. Future large-scale studies that assess the effect of outpatient antimicrobial stewardship and clinically relevant outcomes, including antimicrobial resistance, are needed. In the interim, the growing threat from antimicrobial resistance combined with the available evidence supporting outpatient ASPs makes a compelling argument for the widespread implementation of such programs, even as we await further data.
Acknowledgments
This article is based on research conducted by the Minneapolis Evidence-based Synthesis Program and funded by the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development, Quality Enhancement Research Initiative. The findings and conclusions in this document are those of the authors who are responsible for its contents; the findings and conclusions do not necessarily represent the views of the Department of Veterans Affairs or the United States government.
We thank members of a technical expert panel (Kelly Echevarria, PharmD; Matthew Goetz, MD; Christopher Graber, MD; Allison Kelly, MD, MSOH; Melinda Neuhauser, PharmD, MPH; Gary Roselle, MD) and peer reviewers (Sylvain DeLisle, MD; Graeme Forrest, MD; Chris Gentry, PharmD) of the evidence report for providing advice and feedback. Technical expert panel members and peer reviewers were not compensated for their contributions.
Financial support: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript, and decision to submit the manuscript for publication.
Potential conflicts of interest: None reported.
Disclaimer: The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States Government.
Supplementary Materials
To view Supplementary Materials for this article, please visit http://dx.doi.org/10.1017/ice.2014.41
APPENDIX
Search Strategy
Database: Ovid MEDLINE(R)
1. antibiot$.mp. or exp antibiotics/
2. antimicrob$.mp.
3. exp Anti-Bacterial Agents/
4. exp Anti-Infective Agents, Urinary/
5. exp Cross Infection/
6. exp Community-Acquired Infections/
7. exp Respiratory Tract Infections/
8. exp Wound Infection/
9. exp Catheter-Related Infections/
10. exp Vancomycin Resistance/ or exp Vancomycin/ or vancomycin.mp.
11. aminoglycosides.mp. or exp Aminoglycosides/
12. fluoroquinolones.mp. or exp Fluoroquinolones/
13. broad spectrum antibiotics.mp.
14. carbapenems.mp. or exp Carbapenems/
15. exp Cephalosporins/or broad spectrum cephalosporins.mp.
16. or/1-15
17. exp Education/or education.mp.
18. information campaign.mp.
19. audit.mp.
20. feedback.mp. or exp Feedback/
21. dissemination.mp. or exp Information Dissemination/
22. provider reminders.mp.
23. computerized medical records.mp. or exp Medical Records Systems, Computerized/
24. exp Physician Incentive Plans/ or financial incentives.mp.
25. discharge planning.mp.
26. guideline implementation.mp.
27. guideline adherence.mp. or exp Guideline Adherence/
28. exp Quality Assurance, Health Care/ or quality assurance.mp.
29. program evaluation.mp. or exp Program Evaluation/
30. exp Practice Guideline/
31. exp Physician's Practice Patterns/
32. exp Drug Prescriptions/
33. exp Drug Utilization/
34. or/17-33
35. randomized controlled trial.mp. or exp Randomized Controlled Trial/
36. controlled clinical trial.mp. or exp Controlled Clinical Trial/
37. intervention study.mp. or exp Intervention Studies/
38. Comparative Study/
39. experiment.mp.
40. time series.mp.
41. pre-post test.mp.
42. (randomized controlled trial or controlled clinical trial).pt.
43. (randomized controlled trials or random allocation or clinical trial or double blind method or single blind method).sh.
44. exp clinical trial/
45. (clin$ adj25 trial$).ti,ab.
46. ((singl$ or doubl$ or trebl$ or trip$) adj25 (blind$ or mask$)).ti,ab.
47. (research design or placebos).sh.
48. (placebo$ or random$).ti,ab.
49. exp Double-Blind Method/
50. exp cohort studies/ or (cohort adj (study or studies)).tw. or Cohort analy$.tw. or (Follow up adj (study or studies)).tw. or (observational adj (study or studies)).tw. or Longitudinal.tw. or comparative study/ or follow-up studies/ or prospective studies/ or cohort.mp. or compared.mp. or multivariate.mp. (4148897)
51. (“time series” or pre-post or “Before and after” or intervention).tw.
52. or/35-51
53. 16 and 34 and 52
54. limit 53 to english language
55. limit 54 to humans
56. limit 55 to yr="2000 -Current"
57. (influenza$ or antimalar$ or malaria$ or prophylax$).mp.
58. 56 not 57
