Impact of a prospective audit and feedback antimicrobial stewardship programme on carbapenem consumption: a quasi-experimental study (IPANEMA study)

Abstract

Objectives

To test the hypothesis that a prospective audit and feedback (PAF) intervention combined with electronic tools will reduce carbapenem use without negatively affecting patient outcomes.

Methods

A quasi-experimental, pre-intervention and intervention study was performed conducted in the urology department of a university hospital. The intervention involved implementing a PAF within an antimicrobial stewardship programme with the aid of an electronic tool. The primary outcome was carbapenem use, assessed by DDD/100 patient-days (PD). Secondary outcomes included evaluating the effect of the intervention on overall antibiotic use measured by DDD/100 PD and days of therapy (DOT)/100 PD, as well as patient safety. The chi-squared test or t-test was used, and the Poisson model was employed to assess the association between the intervention and outcomes.

Results

A 9% decrease in carbapenem DDD/100 PD was observed during the intervention period (IR = 0.91; 95% CI = 0.85–0.97, P = 0.007). The proportion of patients who received carbapenem treatment dropped from 17.8% to 16.5% [incidence ratio (IR) = 0.95; 95% CI = 0.86–2.05, P = 0.31]. Carbapenem DOT/100 PD decreased from 12.4 to 11.0 (IR = 0.89; 95% CI = 0.83–0.94, P < 0.001). Overall antibiotic DDD/100 PD decreased by 3% (IR = 0.97; 95% CI = 0.94–0.99, P = 0.001) and DOT/100 PD by 7% (IR = 0.93; 95% CI = 0.91–0.95, P < 0.001). The incidence of infections caused by carbapenemase-producing microorganisms, Enterococcus faecium bacteraemia and Clostridioides difficile-associated diarrhoea episodes was similar in the pre-intervention and intervention periods. ESBL incidence rate decreased, but the differences were not statistically significant (3.94/1000 PD versus 2.88/1000 PD, P = 0.111). Length of hospital stay, in-hospital all-cause mortality, and 30 day readmission incidence remained unchanged.

Conclusions

The implementation of PAF combined with an electronic tool was an effective and safe intervention for reducing carbapenem use.

Introduction

The global increase in bacterial resistance in recent years has been recognized by public health organizations as a priority issue that requires urgent attention from the scientific community.^1–3 Around one-third of hospitalized patients and more than two-thirds of critically ill patients are receiving antimicrobial therapy at any given time. Of particular concern is the fact that half of antibiotic prescriptions are inappropriate or even unnecessary.⁴

Epidemiological studies have demonstrated a direct relationship between antibiotic consumption and the emergence and dissemination of multidrug resistance.⁵ It is important to note that carbapenem overuse has been linked to the emergence of carbapenemase-producing (CP) organisms, which are increasing at an alarming pace worldwide.⁶ Given this concerning public health issue, there is an urgent need to develop effective and safe strategies to promote more judicious use of antibiotics.⁷ Notably, the use of carbapenem-sparing alternatives is currently considered as the cornerstone intervention for avoiding resistance to these antibiotics.⁶

One strategy to reduce antimicrobial consumption is based on the use of hospital-based antimicrobial stewardship programmes (ASPs).⁸ These strategies involve coordinated interventions designed to measure and improve the appropriate use of antibiotic agents by promoting the selection of the optimal drug regimen, including dosing, duration and route of administration.⁹ Prospective audit and feedback (PAF) is an intervention that engages the prescriber after initiation of antibiotic treatment and is a core component of any ASP.¹⁰ Its efficiency is likely to be improved by the use of innovative information technology, which is destined to play an increasing role in ASPs by assisting in patient screening and providing pertinent real-time data in an easily accessible location.^11,12

Several studies comparing antimicrobial drug use before and after the introduction of an ASP have reported reductions in drug consumption and related costs due to the elimination of inappropriate prescribing practices and the shortening of hospital stays.^13,14 For its part, PAF has been shown to be useful in reducing broad-spectrum antibiotic use, antibiotic-resistant Enterobacteriaceae, and Clostridioides difficile-associated diarrhoea (CDD).^15,16 Although the existing literature highlights the importance of PAF in combination with electronic tools, its impact probably depends on how frequently it is completed, how quickly antibiotic prescription is initiated afterwards, and which antibiotics are targeted.¹⁷

The Catalan Health Institute (Catalan acronym ICS), the largest public health service provider in Catalonia, offers care to nearly 6 million people. Interestingly, the ICS has developed an online real-time tool (the SAP business object, SAP-BO), which allows for the selection and analysis of the antibiotic treatments administered to all hospitalized patients, and measures improvements in quality achieved by the implementation of an ASP. The tool also facilitates risk-adjusted inter- and intra-facility benchmarking of antimicrobial usage and evaluates trends in antimicrobial usage over time at particular facilities and at national level as well.

The present study aims to test the hypothesis that a PAF intervention using the SAP-BO tool will reduce carbapenem use without negatively impacting patient outcomes. This strategy may also reduce the frequency of CP Gram-negative microorganism infections, bacteraemia due to Enterococcus faecium, and CDD rates.

Materials and methods

Design

We performed a single-centre, quasi-experimental, pre-intervention and intervention study, which was divided into two periods based on the antibiotic management policy in place. The pre-intervention period comprised retrospective data collection from 1 January 2016 to 31 December 2017, while the intervention period comprised prospective data collection from 1 January 2018 to 31 December 2019.

Setting

The study was conducted at the urology department of Bellvitge University Hospital in Barcelona, Spain. This department was selected due to its high carbapenem consumption rate, as demonstrated in our previous years’ records, and the increasing number of infections caused by CP Gram-negative microorganisms. The ASP team comprised two infectious disease physicians, a microbiologist and two hospital pharmacists.

Patient cohorts

The pre-intervention cohort comprised adult patients (> 17 years old) admitted to the urology department from 1 January 2016 to 31 December 2017. From 1 January 2018 until 31 December 2019, patients who met the eligibility criteria were included in the intervention cohort.

Patients eligible for inclusion in the intervention cohort were those receiving carbapenem treatment, regardless of the duration of treatment, or those receiving any other non-carbapenem treatments for more than 6 days. There were no exclusion criteria.

Intervention

SAP-BO was used for PAF implementation. This electronic screening tool was designed and developed by the ICS, and became available for use on 1 January 2018. It allows real-time viewing, sorting and analysis of the treatment regimens of all hospitalized patients. Specifically, the SAP-BO software is integrated with the electronic medical and nursing records providing on-time screening according to different criteria specified by the user, such as type of antibiotic, treatment duration and inpatient service. It also provides data on antibiotic consumption, which allows for close monitoring and helps to guide antimicrobial stewards in assessing their programme intervention. Additionally, it reflects the measure of success of an ASP intervention. The indicators provided by SAP-BO are DDD and days of therapy (DOT) standardized per 100 patient-days (PD), the proportion of patients under antibiotic treatment, and those receiving targeted antimicrobials such as carbapenems, cephalosporins, quinolones and broad-spectrum antibiotics. All these data can be analysed by time periods.

During the intervention period, SAP-BO reports were generated every weekday at 08:00, providing a list of patients admitted who met the inclusion criteria. The initial evaluation of each included patient was performed by clinical pharmacists (M.R., A.P.), based on clinical situation and infectious-related data such as site of infection, microorganisms isolated and resistance pattern. All pharmacists’ recommendations for optimization and challenging cases were reviewed daily with infectious diseases physicians (I.G., J.C.) and a microbiologist (F.T.). Following this discussion, a note was placed in the patient’s electronic medical record, providing specific recommendations regarding antibiotic therapy optimization. The decision to change an antibiotic was left to the clinical team caring for the patient. No other antimicrobial stewardship strategies were applied.

Outcomes

The primary outcome of the study was the use of carbapenem among patients admitted to the urology department during the pre-intervention and intervention periods. Carbapenem use was assessed using DDDs, which were based on the nursing administration records of the electronic prescription programme. The DDDs were calculated using the Anatomical Therapeutic Chemical Index of the WHO Centre for Drugs Statistics Methodology and were standardized by 100 PD.

Secondary outcomes were:

• The percentage of patients under carbapenem treatment assessed quarterly. The proportion was the ratio between the number of patients under carbapenem treatment and the number of patients under antibiotic treatment.

• Carbapenem DOT standardized by 100 PD.

• Overall antibiotics DDD standardized by 100 PD.

• Overall antibiotics DOT standardized by 100 PD.

• Infections due to CP Gram-negative microorganisms defined as the number of cases per year per 1000 PD, including CP Pseudomonas aeruginosa, CP Enterobacteriaceae and both ESBL-producing P. aeruginosa and Enterobacteriaceae. Patients could be included as two independent episodes if two different resistant microorganisms were isolated in different samples. Susceptibility data were obtained from the microbiology department. MIC breakpoints were defined in compliance with the CLSI documents.

• Bacteraemia due to E. faecium defined as the number of cases per year per 1000 PD.

• Hospital-acquired CDD defined as the number of cases per year per 1000 PD.

• Median length of hospital stay.

• Incidence of readmission within 30 days pertaining to an emergency department visit or admission to an inpatient unit.

• In-hospital all-cause mortality.

The study outcomes were assessed by at least two members of the ASP team.

Ethics

This study was evaluated and approved by the Clinical Research Ethics Committee of Bellvitge University Hospital (PR340/22). Written informed consent was not deemed necessary. Patients’ data were de-identified and coded to ensure privacy. No patient identifiers were recorded or transcribed into the database for analysis. The confidentiality of the patients’ data was protected in accordance with the Declaration of Helsinki and national and institutional standards.

Statistical analysis

The demographic profiles and safety outcomes of the subjects included were described according to the year and the patient cohorts (2016–17 versus 2018–19). Counts and percentages were presented for categorical variables, and means with standard deviation (SD) or the medians with IQRs were presented for numerical variables.

Chi-squared tests were performed for binary variables to assess differences between the pre-intervention and intervention groups. The t-test was used for continuous normal variables, while Wilcoxon signed-rank tests were used for continuous non-normal variables.

To assess the association of DDD and DOT with the study group, Poisson models were used, taking the number of stays as the offset of the model. A beta regression was performed for the association of the percentage of carbapenems. The effect of the year and the trimester were also included. Incidence ratios (IRs) of the estimated effects for Poisson models and estimates for beta models are presented in tables.

The conditions of application of the models were validated, and the 95% CIs of the estimators were calculated whenever possible. All analyses were performed using the statistical package R version 4.1.0 (2021-05-18) for Windows.

Results

Population

A total of 7631 patients were included in the study, with 4092 in the pre-intervention period (54%) and 3539 in the intervention period (46%). Table 1 shows demographic and clinical variables. Differences were noted in age and patient complexity, but they were not clinically relevant.

Primary outcome

There was a 9% decrease in carbapenem use assessed by DDD/100 PD in the intervention period (IR 0.91; 95% CI = 0.85–0.97, P = 0.007). After adjusting the model for trimester, we estimated an 8% reduction in carbapenem consumption (IR 0.92; 95% CI = 0.85–0.98, P = 0.013) (Figure 1a). Trimester carbapenem consumption evolution per trimester is shown in Figure S1a (available as Supplementary data at JAC Online).

Secondary outcomes

In the intervention period, the proportion of patients who received carbapenem treatment decreased from 17.8% (388/2179) to 16.5% (294/1777), but the difference was not statistically significant (IR 0.95; 95% CI = 0.86–1.05, P = 0.31) (Figure 1b, Figure S1b). Carbapenem DOT/100 PD also presented a downward trend, decreasing from 12.4 to 11.0 (−1.5 DOT/100 PD, IR 0.89; 95% CI = 0.83–0.94, P < 0.001). In the model adjusted by trimester, the intervention DOT/100 PD value was also 11% less than in the pre-intervention period (IR 0.89; 95% CI = 0.84–0.95, P < 0.001) (Figure 1c, Figure S1c). Overall antibiotic use demonstrated a significant change in trend after the ASP intervention, with DDD/100 PD decreasing by 3% (IR 0.97; 95% CI = 0.94–0.99, P = 0.002) and the DOT/100 PD decreasing by 7% (IR 0.93; 95% CI = 0.91–0.95, P < 0.001).

The incidence of infections by CP microorganisms did not significantly change before and after the intervention (1.21 versus 1.57 cases/1000 PD, P = 0.451). There were also no significant changes in the number of E. faecium invasive isolates (0.11 versus 0.06 cases/1000 PD, P = 1) or the CDD episodes (0.47 versus 0.56 cases/1000 PD, P = 0.892). Conversely, the incidence of infections caused by ESBL microorganisms decreased from 3.94 cases/1000 PD (75 cases) in the pre-intervention period to 2.88 cases/1000 PD (46 cases) in the intervention period; however, the difference did not reach statistical significance (P = 0.111).

Median length of hospital stay remained unchanged after the intervention (3 days in both periods). The in-hospital all-cause mortality was not significantly different between the pre-intervention and intervention periods (9 versus 14 deaths, P = 0.235), and the incidence of 30 day readmission was also similar between periods (11.8 versus 11.8 patient readmissions, P = 0.278) (Table 2).

Discussion

Our study demonstrates that the implementation of a PAF intervention combined with an electronic tool successfully reduced carbapenem use, as assessed by DDD/100 PD and DOT/100 PD without compromising patient safety. Additionally, the overall antibiotic use was also reduced, suggesting that the intervention was effective and did not trigger the previously described phenomenon of ‘squeezing the balloon’, whereby the increased use of other antibiotic offsets any reduction in carbapenem use.

However, it is worth noting that our intervention had no impact on the number of patients who initiated carbapenem treatment. While we did not implement restrictive measures, our primary interventions involved de-escalating to narrower-spectrum antibiotics and shortening treatment durations, which may account for our findings.

Lopez-Viñau et al. were able to achieve an 80.46% reduction in CP Gram-negative microorganisms after 2 years of implementing an ASP that included a bundle of educational and restrictive measures.¹⁸ In another study, Komatsu et al. reported a reduction in antibiotic consumption without any changes in P. aeruginosa antibiotic resistance pattern.¹⁹ However, in our study we did not observe a reduction in infections caused by CP microorganisms. While some studies have shown a direct relationship between the reduction of in-hospital antibiotic consumption and a decrease in ESBL isolates,^20,21 there is no clear evidence of a direct relationship between carbapenem consumption and antimicrobial resistance.

Both the incidence of E. faecium bacteraemia and CDD episodes remained low and stable, as reported by García-Rodríguez et al. after 4 years of implementing an ASP focused on carbapenem consumption.²² The prevalence of antimicrobial resistance is a result of multiple factors, including the use of infection control bundles and indicators of hospital complexity, which remained unchanged during the study period.

In our study, immediate results were not observed following the intervention, and the reduction in carbapenem use was not observed until the third quarter of 2018, and did not have a statistically significant ecological effect. Other studies have reported that implementing an ASP for a period shorter than 12–24 months is insufficient to identify the true impact of the programme on antimicrobial resistance.²³ This suggests that longer ASPs may be needed to detect changes in local susceptibility patterns of microorganisms.

ASPs should focus their efforts on optimizing antibiotic treatments in wards with high consumption of broad-spectrum antibiotics.²⁴ We decided to focus on carbapenems because their use in the urology department has increased in recent years, and they have the broadest antibacterial spectrum and the ability to rapidly induce β-lactamases, including carbapenemases.⁶ In addition, ASPs are generally underfunded, so it is necessary to prioritize highly efficient interventions.

In our study, the reduction in carbapenem use had no effect on mortality or readmission rates. Other studies have also reported that carbapenem-targeted programmes do not increase mortality. In this sense, Seah et al. and García-Rodríguez et al. showed that reducing and optimizing the use of carbapenems did not compromise patients’ safety in terms of clinical outcomes.^22,25

In the last decade, the number of studies applying a decision support system for antibiotic treatment has increased. In 2013, Leivobici et al. published a controlled trial in which a computerized decision support system selected the empirical antibiotic treatment without reducing the long-term mortality of patients.²⁶ Another recent study of an ASP based on computerized decision-based systems was published by Catho et al.; however, they did not find differences in DOTs when applying an electronic tool-based PAF in contrast to the standard ASP.²⁷ In our study, we used an electronic support system for selecting patients in whom ASP may be applied, but the decisions regarding antibiotic optimization were made by physicians.

Our study has several limitations. First, its single-centre, quasi-experimental design and the fact that the intervention was applied solely to a urology department, rather than the entire hospital, may limit the generalizability of the results to other settings. Second, no cost-effectiveness analysis was performed. Finally, although we observed a decrease in the incidence of ESBL infections, we did not observe a decrease in CP infections; however, it is important to note that our study was not designed to establish causal relationships between drug use and antibiotic resistance.

Overall, our study demonstrates the usefulness and safety of the implementation of an ASP that combines a PAF intervention with an electronic tool to reduce carbapenem use. Further studies are needed to explore the ecological impact of similar programmes.

Funding

This work was supported by CIBERINFEC (CB21/13/00 009) and CIBERES (CB06/06/0037), Instituto de Salud Carlos III, Madrid, Spain.

Transparency declarations

None of the authors have any conflicts of interest to declare. A.P. and J.C. were responsible for the conception and design of the work; M.R., I.G. and F.T. were in charge of data acquisition, and A.P., J.C., M.R., E.S., F.V. and P.S. analysed and interpreted the data. All authors approved the final version.

Supplementary data

Figure S1 is available as Supplementary data at JAC Online.

References