Interventions to improve antibiotic prescribing practices for hospital inpatients
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
The primary aim is to systematically review the literature to identify interventions that alone, or in combination, are effective in improving antibiotic prescribing to hospital inpatients.
Objectives
1. To estimate the effect of interventions on three aspects of antibiotic prescribing:
1.a: The decision to prescribe an antibiotic for prophylaxis or treatment;
1.b: The antibiotic regimen: drug(s) selected, dose and route of administration;
1.c: The duration of prophylaxis or treatment.
2. To compare the effectiveness of a combination of interventions versus a single intervention.
3. To estimate the effect of increasing appropriate, evidence‐based antibiotic prescribing on patient outcomes and healthcare costs. Specific outcomes will be:
3.a: Antibiotic prescribing and costs;
3.b: Other healthcare costs (e.g. investigations, length of hospital stay, readmission to hospital);
3.b. Antibiotic associated diarrhoea and Clostridium difficile associated diarrhoea;
3.c. Prevalence of colonisation by or clinical infection by antibiotic resistant bacteria;
3.d: Other measures of clinical outcome (e.g. 28 day or in‐hospital mortality).
Background
Antibiotic resistance is now regarded as a major public health problem. In comparison with infections caused by susceptible bacteria, infections caused by multidrug‐resistant bacteria are associated with higher mortality, morbidity and prolonged hospital stay, which is at least twice as long for patients with community‐ or hospital‐acquired infections caused by resistant pathogens (Holmberg 1987). The emergence of multidrug‐resistant organisms limits the choice of therapy for patients with hospital‐acquired infections and, ominously, for the first time since antibiotics were introduced, we are faced with the prospect of not having any effective chemotherapy for patients with bacterial infections.
Despite strenuous efforts to control antibiotic usage and to promote optimal prescribing, practitioners continue to prescribe excessively and inappropriately; it is estimated that up to 50% of antibiotic usage in hospitals is inappropriate (House of Lords 1998). Recently, a number of reports have proposed a range of measures designed to address the problem of increasing resistance (Copenhagen 1998; House of Lords 1998; House of Lords 2001; SMACS 1998). Common to all the recommendations is the challenge to reduce inappropriate and excessive antibiotic prescribing, the implication being that antibiotic resistance is largely a consequence of the selective pressures of antibiotic usage and that reducing these pressures by the judicious administration of antibiotics will facilitate a return of susceptible bacteria or, at least, will prevent or slow the pace of the emergence of resistant strains.
The Cochrane Database of Systematic Reviews contains seven systematic reviews of evidence about the effectiveness of six different methods for changing practice: audit and feedback of information (Thomson 1998a; Thomson 1998b); computerised advice on drug dosage (Walton 2001); educational outreach visits (Thomson 1997); local opinion leaders (Thomson 1999); mass media interventions (Grilli 1998) and printed educational materials (Freemantle 1996). There are two reviews about audit and feedback: one deals with this intervention alone (Thomson 1998b) and the second with complementary interventions to enhance its effectiveness (Thomson 1998a). Each of the seven reviews includes studies that are directly relevant to antibiotic policies or to improvement in professional practice for the management of infection. However, while each of the reviews shows that all six of these interventions can be effective, overall they provide little evidence about the relative effectiveness of the different methods in specific contexts or about the importance of other potential effect modifiers, such as content, source, timing, recipient and format.
This review of interventions to improve prescribing of antibiotics to hospital inpatients will be proceeding in parallel with a review of interventions to improve prescribing of antibiotics to patients in ambulatory care (Arnold 2002).
The literature contains many examples of control of outbreaks of infection with drug resistant bacteria that may be attributed to antibiotic control measures. However, it has been very difficult to eliminate confounding variables, such as other infection control measures or change in patient case mix (McGowan 1983). The methodology of earlier studies was crude, both in terms of microbiology and epidemiology. More recent studies have used molecular techniques for precise definition of cross‐infection and more rigorous case control to minimise the impact of confounding variables (Richard 1994b; Richard 1994a). Association between prescribing and resistance is strengthened if the relationship can be demonstrated across a range of institutions. A study conducted in 18 centres in the United States identified a significant correlation between ceftazidime prescribing in intensive care units and the prevalence of ceftazidime resistant Enterobacter cloacae (Ballow 1992). However, the best evidence comes from studies that relate changes in antibiotic prescribing to changes in resistance (Bradley 1999; de Man 2000).
Colitis and diarrhoea caused by Clostridium difficile are important hospital acquired infections that are caused by antibiotic prescribing (Bartlett 1992; Begley 1994; Fekety 1995). Interventions to change antibiotic prescribing can reduce the prevalence of Clostridium difficile‐associated diarrhoea (Climo 1998; Ludlam 1999; McNulty 1997).
Placebo controlled trials have established the effectiveness of surgical antibiotic prophylaxis for reducing risk of wound infection (SIGN 2000) and observational studies consistently show an association between prompt initiation of therapeutic antibiotics for bacteraemia or other serious infections and clinical outcome (Gleason 1999; Gransden 1990;Ispahani 1987; Pedersen 1997). It is therefore logical to expect to find evidence that implementing evidence‐based guidelines about antibiotic prophylaxis or treatment improves clinical outcome (Nathwani 2000).
Objectives
The primary aim is to systematically review the literature to identify interventions that alone, or in combination, are effective in improving antibiotic prescribing to hospital inpatients.
Objectives
1. To estimate the effect of interventions on three aspects of antibiotic prescribing:
1.a: The decision to prescribe an antibiotic for prophylaxis or treatment;
1.b: The antibiotic regimen: drug(s) selected, dose and route of administration;
1.c: The duration of prophylaxis or treatment.
2. To compare the effectiveness of a combination of interventions versus a single intervention.
3. To estimate the effect of increasing appropriate, evidence‐based antibiotic prescribing on patient outcomes and healthcare costs. Specific outcomes will be:
3.a: Antibiotic prescribing and costs;
3.b: Other healthcare costs (e.g. investigations, length of hospital stay, readmission to hospital);
3.b. Antibiotic associated diarrhoea and Clostridium difficile associated diarrhoea;
3.c. Prevalence of colonisation by or clinical infection by antibiotic resistant bacteria;
3.d: Other measures of clinical outcome (e.g. 28 day or in‐hospital mortality).
Methods
Criteria for considering studies for this review
Types of studies
All randomised and quasi‐randomised controlled trials (RCTs/CCTs), controlled before and after studies (CBAs) and interrupted time series (ITSs) studies will be included.
Types of participants
Health care professionals who prescribe antibiotics to hospital inpatients receiving acute care. The review will exclude interventions targeted at residents in nursing homes or other longterm healthcare settings.
Types of interventions
The following professional interventions in the Effective Practice and Organisation of Care Group (EPOC) scope will be included:
1. Educational interventions: distribution of educational materials; educational meetings, local consensus processes, educational outreach visits, local opinion leaders, reminders provided verbally, on paper or on computer.
2. Patient mediated interventions: new clinical information (not previously available) collected directly from patients and given to the professional. Examples include scores for the assessment of severity of community acquired pneumonia or of the probability of bacterial infection in a patient with post‐operative fever.
3. Audit and feedback.
4. Restrictive interventions. These include selective reporting of laboratory susceptibilities, formulary restriction; requiring prior authorisation of prescriptions by infectious diseases physicians, microbiologists, pharmacists etc.; therapeutic substitution and automatic stop orders.
5. Antibiotic policy change strategies including cycling, rotation and crossover studies.
6. Financial: organisational or professional financial incentives for antibiotic prescribing.
7. Organisational: creation of new teams for influencing or supervising antibiotic prescribing.
8. Structural: the influence on antibiotic prescribing of changing from paper to computerised records and of the introduction or organisation of quality monitoring mechanisms.
Excluded studies
1. Descriptions of interventions to change antibiotic prescribing without measurement of the effect of these interventions on prescribing or other outcome measures.
2. Surveys of hospitals to establish the range of measures used to control or optimise antibiotic prescribing.
Types of outcome measures
The primary outcome measure will be the rate of appropriate antibiotic prescribing. This will include one or more of the following: the decision to prescribe an antibiotic or the prescribing of a recommended choice, dose or duration of antibiotic(s) for specific conditions. Data extraction will include information about the evidence supporting each study's definition of appropriate prescribing (e.g. practice guidelines assessed with MERGE criteria).
Secondary outcome measures will be:
1. Healthcare resource costs.
2. Indicators of the prevalence of antibiotic resistant bacteria or Clostridium difficile (either cases of suspected clinical infection or prevalence of asymptomatic carriage).
3. Indicators of clinical outcome such as length of hospital stay, inpatient mortality and 28 day mortality.
Studies examining prescribing of multiple drug classes will be included provided that specific data on antibiotic prescribing can be extracted.
Search methods for identification of studies
The literature search will make use of two existing databases (Literature searches 1 and 2) and will repeat the same search strategies using the Cochrane Effective Practice and Organisation of Care Group (EPOC) specialised register (Literature search 3). The results of the three literature searches will be compared.
1. Literature search 1: conducted by Erwin Brown and Giles Hartman.
1.1.Medline was searched through PubMed (thus circumventing the need for MESH terms) from 1980 onwards with the search terms: "antibiotic use AND resistance", "Antibiotic guidelines", "antibiotic guidelines AND implementation", "antibiotic policies", "antibiotic policies AND prescribing" "antibiotic policies AND antibiotic use" ...all of these search terms were then re‐used with "antimicrobial" instead of "antibiotic"; "optimal antibiotic/antimicrobial prescribing/use".
1.2. The Cochrane Library was similarly searched with these search terms with the assistance of Robin Harbour from the SIGN methodology group.
1.3.EMBASE was searched from 1988 to the present with the same search terms.
1.4. The list of references thus generated included around 300 about control of antibiotic prescribing in hospitals. These were then manually searched and quoted references not identified by electronic means were added. This brought the total to over 529. As an internal check each new reference acquired was searched and any previously unidentified paper was added to the database.
2. Literature search 2: conducted by Peter Gross and Duressa Pujat for a systematic review: "Implementing Practice Guidelines for Appropriate Antimicrobial Usage."
2.1 This search was conducted in MEDLINE from 1966 to December 12 2000 using PubMed and OVID. The Cochrane Library was also searched.
2.2 The search used the exploded term "antibiotics" and "premedication" which included "antibiotic prophylaxis". The new term "guideline adherence" expressed most closely the concept of "guideline implementation" but it is only available for articles indexed from 1998. Therefore for the "guideline" concept the additional MESH terms "guidelines" and "practice guidelines" were used. The MESH terms "clinical protocols", "critical pathways", "evidence based medicine" and "guideline" were used as textwords. To capture the concept of "implementation" the search used the MESH terms "physicians practice patterns", "knowledge, attitudes, practice", "attitude of health personnel" and "prescription, drugs" plus the truncated textwords: implement recommend and adhere.
2.3 To augment the retrieval the following MESH headings were combined with the search terms in 2.2: "cross infection", "community acquired infection" and the MESH terms for respiratory tract infections, wound infections and catheter associated infections.
3. Literature search 3: the EPOC specialised register will be searched for relevant studies using the same terms as Literature Searches 1 and 2.
3.1.The EPOC register has been compiled by searching MEDLINE (back to 1966), Health STAR (back to 1975), and EMBASE (back to 1980). The search strategy used has a sensitivity of 92.4% compared to the gold standard. The gold standard for comparison are those studies found by hand searching Medical Care (1969‐95), BMJ (1992‐94) and full text searching of OVID Biomedical Core Collection of all original and miscellaneous article from Annals of Internal Medicine, BMJ, JAMA and Lancet (1995‐96).
3.2. Additional studies will be sought by searching the bibliographies of included articles, the Science Citation Index for included studies, personal files and by contacting experts in the area regarding any unpublished work.
There will be no language limitations for the literature review.
Data collection and analysis
Erwin Brown and Giles Hartman will review citations and abstracts retrieved in the search to identify all reports that include original data about interventions to change antibiotic prescribing. Peter Davey and Giles Hartman will assess the methodological quality of included studies according to the criteria developed by the EPOC group (EPOC Data Collection Checklist and Draft EPOC Methods Paper on Including Interrupted Time Series Designs in a EPOC Review)/
Assessments for inclusion will be done independently by Peter Davey and Giles Hartman, without blinding to study author or location. Disagreements were resolved by discussion and consensus.
Data abstraction will be performed by two independent reviewers using a template including information on: study design, type of intervention, presence of controls, type of targeted behaviour, participants, setting, methods (unit of allocation, unit of analysis, study power, methodological quality, consumer involvement), outcomes, and results. Data from RCTs and CBAs will be presented using the format suggested in the EPOC Working Paper on presentation of data.
Data will be reported in natural units in the results table and the results section. Where possible the dichotomous variables (appropriate antibiotic prescription or not, as determined by study authors) will be reported as relative risk and risk difference with 95% confidence limits. In order to capture differences between baseline and post‐intervention measurements, when results include a continuous variable, the raw mean percent change in appropriate prescribing with 95% confidence limits and p‐values will be reported. Where trials compare more than one intervention or combination of interventions, each comparison will be handled separately.
It is likely that the review will identify studies with unit of analysis errors (for example, randomisation by hospital with analysis by patient without any adjustment for clustering). The results of these studies will be presented as point estimates of the intervention effect without presentation of any statistical analysis or confidence intervals.
Methods of analysis of Interrupted Time Series data will be examined critically. The preferred method will be a statistical comparison of time trends before and after the intervention. If the original paper does not include an analysis of this type the data presented will be used to perform new analyses with the recommended EPOC methods.
The retrieved studies are likely to be heterogeneous with respect to study design, intervention and study subjects and completeness of reporting of results. Studies will be reviewed qualitatively and, where appropriate, quantitative sub‐group analysis by similar intervention will be performed using a random effects model. Studies that involve multiple interventions will be combined statistically if the interventions are sufficiently similar. Sensitivity analysis for pooled results will be performed based on methodological quality, if possible.
