Qual Saf Health Care 17:387-392 doi:10.1136/qshc.2007.023267
  • Quality improvement report

Clinical and economic impact of an antibiotics stewardship programme in a regional hospital in Hong Kong

  1. C K Ng1,
  2. T C Wu1,
  3. W M J Chan1,
  4. Y S W Leung2,
  5. C K P Li1,
  6. D N C Tsang3,
  7. G M Leung4
  1. 1
    Department of Medicine, Queen Elizabeth Hospital, Hong Kong, China
  2. 2
    Department of Pharmacy, Queen Elizabeth Hospital, Hong Kong, China
  3. 3
    Department of Pathology, Queen Elizabeth Hospital, Hong Kong, China
  4. 4
    School of Public Health, Li Ka Shing Faculty of Medicine, University of Hong Kong, Hong Kong, China
  1. Dr C K Ng, Department of Medicine, Queen Elizabeth Hospital, 30 Gascoigne Road, Hong Kong, China; ngck6{at}
  • Accepted 12 November 2007


Background: Inappropriate use of antibiotics is one of the important factors attributing to emergence of drug-resistant pathogens. Infection with multidrug-resistant pathogens adversely affects quality of medical care.

Context: Queen Elizabeth Hospital, an 1800-bed acute service hospital in Hong Kong. Antibiotics are commonly prescribed for treating acute infections.

Key measures for improvement: Reduce inappropriate prescription of broad-spectrum antibiotics and overall antibiotic prescription through implementation of a multidisciplinary antibiotics stewardship programme (ASP).

Strategies for change: A multidisciplinary programme involving policy and guideline formulation, education and feedback, monthly antibiotic consumption and cost monitoring, antimicrobial susceptibility pattern reporting and concurrent feedbacks for commonly prescribed broad-spectrum antibiotics was implemented in 2004. Predefined logistics to prescribe “restricted” antibiotics were formulated and implemented with collaborative efforts from clinical and non-clinical departments. The programme was supported by management at department and hospital levels.

Effects of change: Broad-spectrum antibiotics were prescribed inappropriately in 28.9% (n = 192) clinical scenarios. The ASP reduced the restricted and total antibiotic consumption as well as the antibiotics-related costs. Predefined clinical outcomes were not adversely affected. Economic analysis suggested that the extra human cost in running ASP could be offset by savings from antibiotic expenditure.

Lessons learned: It is cost-effective to implement a multidisciplinary ASP in acute service hospitals as the programme reduces antibiotic consumption and results in overall cost savings. The quality of medical care is not jeopardised as the important clinical outcomes are not adversely affected. The generalisability and sustainability of ASPs in other clinical contexts warrant further studies to ensure the continuous success of this programme.

The prevalence of nosocomial infections caused by antibiotic-resistant pathogens is escalating worldwide.1 As empirical antibiotics are usually ineffective in controlling infections by resistant clones, the quality of patient care can be adversely affected due to delayed prescription of appropriate life-saving antibiotics, prolonged hospitalisation and increased morbidity and mortality.23 Antibiotic-related expenditure and total healthcare costs are inflated because of the need to resort to second-line, more expensive and possibly more toxic antibiotics and the consequent need to manage the associated side effects.45

The emergence and dissemination of multiresistant microorganisms are driven by complex and inter-related factors: host biology, pathogenicity of infectious agents, pharmacodynamics and the external environment, viz. infection control and microbial ecology.6 Underpinning several of these factors, the inappropriate use of antibiotics is a critically important yet modifiable contributing factor.7 Studies have shown that as much as 50% of antimicrobial use in humans may be inappropriate.89 Overuse and misuse exert perverse selective pressures, favouring the survival and multiplication of resistant clones.1011


Queen Elizabeth Hospital (QEH) is an 1800-bed regional hospital providing acute care service for a catchment population of 0.49 million in the central Kowloon region in Hong Kong. Past overseas experience had suggested that an antibiotics stewardship programme (ASP) can be an effective strategy in combating the emergence of drug resistant pathogens. ASP is defined as the selection of an optimal agent, with an appropriate dose and duration of antimicrobial treatment to ensure the best clinical outcome for the treatment or prevention of infection, with minimal toxicity to the patient and minimal impact on subsequent resistance.1213 However, actual implementation of ASP in clinical practice often faces considerable challenges and resistance from frontline staff. Success is highly contextually specific and the experience at one location may not apply elsewhere. There are fears that ASP could delay prescription of “potent” antibiotics, increase the risk of developing septicaemia, increase admissions to intensive care unit, prolong hospitalisation and increase mortality. Financial consideration is another obstacle to implementing ASP as the human cost related to running the programme is not trivial. The feasibility of implementing ASP, the economic cost associated with ASP, the clinical impact, and cost-effectiveness of ASP in local hospitals has not been well studied before.


A multicomponent, multidisciplinary ASP was introduced in July 2004 as a continuous quality improvement (CQI) programme in the department of medicine, QEH, in collaboration with the department of pathology (microbiology division) and pharmacy. The ASP involved 16 acute medical wards.

The antibiotics CQI programme used different intervention strategies based on the Inter-hospital Multi-disciplinary Programme on Antimicrobial ChemoTherapy (IMPACT) guidelines.14 IMPACT is a local antibiotic guideline that describes the prevalence of antimicrobial resistance in Hong Kong and recommends the optimal choices of antibiotics in treating infections. The chapter on ASPs summarises the advantages of ASP and outlines the impediments during its implementation. Lectures were conducted to update doctors on the latest developments on appropriate antibiotic prescriptions. Concurrent feedback and advice on antibiotics prescription were offered by the infectious disease (ID) team during office hours. Close monitoring of monthly consumption and cost of antimicrobials was implemented with the help of pharmacists. The pharmacists also collected the antibiotics order form (AOF; see web-only supplementary data) and controlled the prescriptions of restricted antibiotics if the AOF was not appropriately filled. Antimicrobial susceptibility patterns were prepared quarterly by microbiologists and circulated to frontline medical staff to alert them of the severity of antibiotic resistance and the most appropriate choice of antibiotics for different pathogens. “Selective reporting” of susceptibilities to antibiotics of which prescription was being encouraged, was implemented to facilitate rotation of antibiotics prescription.15

The principal clinical intervention of the ASP was the restriction of prescription of 12, mainly broad-spectrum, antimicrobials: the anti-pseudomonal cephalosporins (ceftazidime, cefepime, piperacillin-tazobactam, cefoperazone-sulbactam), carbapenems (meropenem, imipenem), intravenous vancomycin, intravenous fluoroquinolones (levofloxacin, ciprofloxacin), intravenous macrolides (azithromycin, clarithromycin) and the antifungal agent fluconazole. The reasons for restricting the use of these classes of antimicrobials included:

  • many of these are commonly used, potent, broad-spectrum antibiotics;

  • the appropriateness of prescription was not well considered;

  • their overuse and misuse might be associated with development of multiresistant pathogens;

  • similar bioavailabilities in oral and intravenous preparations.

Appropriate logistics and workflow for prescription-restricted antibiotics were formulated. Before prescribing any of the restricted antimicrobials, the doctor in charge had to fill in an AOF, documenting the reasons for the choice. The AOF would be sent to the pharmacy to ensure a supply of restricted antibiotics. A reminder letter would be sent to the doctor in charge to fill in the necessary information if the AOF was either not available or had not been duly completed. If a completed AOF was not available, only a 1-day course of the restricted antibiotic was supplied. On receipt of the AOF, a designated infection disease (ID) specialist would review the appropriateness of the prescription. If the prescription deviated from the recommended practice, concurrent feedback would be offered and appropriate, evidence-based alternatives suggested. Such arrangements for the supply of restricted antibiotics prevailed over weekends and public holidays. The ASP team consisted of physicians, a microbiologist, an ID specialist and a pharmacist.

Administrative support in terms of manpower redeployment and workflow rearrangement was facilitated by the chief of service of the department of medicine and consultant microbiologist from department of pathology (microbiology division), QEH. The ASP was approved by the hospital drug and therapeutic committee and the hospital’s chief executive.


Study design

We used a pretest–post-test analysis, comparing the 1-year pre-ASP period (from July 2003 to June 2004) to the 1-year post-ASP period (from July 2004 to June 2005). The impact of ASP on antibiotic consumption, predefined clinical outcomes and antibiotic expenditures was measured and evaluated. The economic assessment was a comparative cost analysis of all costs related to the ASP.

Study population

We included patients admitted to the medical wards between July 2003 and June 2005 who were prescribed any antibiotic. Antibiotics were prescribed in 18 045 admission episodes in the pre-ASP period and in 19 390 admission episodes in the post-ASP period. Patients in the pre-ASP period were slightly younger but had more comorbid conditions on admission (except chronic respiratory and haematological diseases) (table 1).

Table 1 Demographic characteristics of patients. Data are n (%) except age (mean (SD))

Indications for antibiotics

The indications for antibiotic prescriptions are summarised in table 2. Patients in the post-ASP period had more respiratory infections and episodes of septicaemia which essentially required antibiotics. Septicaemia was defined based on the International Classification of Diseases 10th Revision, Clinical Modification (ICD-10).16 Indications for antibiotics could not be confidently identified from the electronic database in 3537 (19.6%) patients in pre-ASP period and 3122 (16.1%) patients in the post-ASP period (p<0.001).

Table 2 Indications for antibiotic prescriptions. Data are n (%)

Appropriateness of restricted antibiotics prescription

To study the extent of inappropriateness of antibiotics prescription, 664 copies of the AOF were randomly selected from all the AOFs reviewed in the initial phase of the ASP. Among these 664 AOFs, we found that restricted antibiotics were prescribed on an empirical basis in 463 (69.7%) patients, for documented infections in 197 (29.6%) patients and as prophylaxis in 4 (0.6%) patients. Third and fourth generation anti-pseudomonal cephalosporins were the most commonly prescribed restricted antibiotics (66.1%). Based on the ID specialist’s assessment, 192 (28.9%) prescriptions were inappropriate. The most common errors were related to the choice of antibiotics (131 episodes, 68.3%), followed by inappropriate route (38 episodes, 20%) and dosage problems (19 episodes, 10%). However, antibiotic prescriptions were considered essential for all those admissions.

Antibiotics consumption

Restricted antibiotics were prescribed in 3537 (19.6%) of admission episodes in 2003–2004 and 2385 (12.3%) of admission episodes in 2004–2005. Antibiotic consumption was measured in defined daily dose or DDD17 and adjusted for patient headcount (DDD/100 patients treated) and patient-days (DDD/1000 patient-days). Consumption of restricted antibiotics showed significant reduction after implementation of ASP (table 3). The median monthly DDD/100 patients treated reduced by 47.2% and the median monthly DDD/1000 patient-days reduced by 43.6%. Significant reduction of non-restricted antibiotics consumption was also observed per 100 patients treated (7.9% reduction, p<0.001) but not for the patient-days outcome (3.8% reduction, p = 0.478) (table 3).

Table 3 Median monthly antibiotics consumption before and after implementation of the antibiotic stewardship programme (ASP)

Clinical outcomes

The mean (SD) lengths of hospital stay before and after implementation of the ASP were 7.46 (10.52) days and 6.97 (9.87) days, respectively (p <0.001). No significant difference in intensive care unit admission episodes was observed between the pre-ASP (464, 2.6%) and post-ASP (454, 2.3%) groups (p = 0.15). Also, there was no statistically significant difference in overall mortality between the pre-ASP group (8.8%) and the post-ASP group (8.4%) (p = 0.28), which was also the case for disease-specific mortality related to infectious disease (3.1% in the pre-ASP group and 3.2% in the post-ASP group, p = 0.55).

Unplanned readmission was defined as readmission within 28 days after discharge. The overall unplanned readmission rate was 17.6% (2897 patients) in the pre-ASP group and 18.7% (3322 patients) in the post-ASP group (p = 0.008). Unplanned readmissions related to infections were significantly higher in the post-ASP group (9.9% vs 11.1%, p <0.001). Multivariate logistic regression showed that presence of septicaemia, having “respiratory disease” as the principal diagnosis and presence of two or more chronic comorbidities were independent predictors for higher unplanned readmissions related to infections (table 4). The multivariate analysis revealed that ASP correlated negatively to unplanned readmissions related to infections.

Table 4 Logistic regression for independent predictors of unplanned readmissions related to infection

Economic assessment

Economic assessment was performed from the hospital perspective.1819 Breakdown of direct and indirect costs incurred by the ASP is summarised in Appendix A. The cost of each item was calculated by multiplying the per unit cost by measured total quantity. Per unit cost was calculated by dividing annual salary by total disposable working hours. Annual total salary was retrieved from the Hospital Authority Notional Annual Mid-point Salary (NAMS) which provided the average salary of a particular rank over that financial year. A positive sign indicated additional expenses and a negative sign indicated savings.

All costs related to ASP are summarised in table 5. Direct ASP costs comprised mainly human costs, which included costs of ID specialists, microbiologists, pharmacists and administrators. The direct ASP cost was estimated to be US$71 294 per year (assuming 1US$ = 7.8HK$). Indirect ASP costs included both human costs and consumable costs and there was savings of US$140 per year.

Table 5 Summary of cost related to the antibiotics stewardship programme (ASP). All costs are in US$

The overall antibiotic savings after implementation of the ASP was US$380 899 per annum. Antibiotics expenditure was adjusted for 100 treated patients and 1000 patient-days (table 5). After implementation of the ASP, the monthly cost of restricted antibiotics per 100 treated patients and per 1000 patient-days were significantly reduced by 52.0% (from US$5605 to US$2690; p<0.001) and by 46.4% (from US$7293 to US$3906; p<0.001), respectively. On the other hand, the cost of non-restricted antibiotics per 100 patients treated and per 1000 patient-days increased by 6.1% (from US$3030 to US$3215; p = 0.052) and 11.9% (from US$3946 to US$4414; p = 0.003), respectively.

The total savings of the ASP was US$309 745 per annum. The cost savings per 100 patients treated per annum was US$2350 and the cost saving per 1000 patient-beds was US$2182 per annum. The average hospital bed cost per patient-day for the department of medicine was US$387.5 per annum during 2003–2005. The ASP cost savings represented 0.56% of the hospital bed costs.

Sensitivity analysis was not performed in this CQI exercise as the savings in antibiotics expenditure (US$380 899) significantly out-weighed the direct and indirect expenses incurred by the programme.


Clinical impact of ASP

One of the desirable targets of an ASP is to control inappropriate and overuse of broad spectrum antibiotics. However, some doctors fear that the ASP might adversely affect the clinical outcomes of their patients since “timely” prescriptions of broad-spectrum antibiotics would be limited. Such a worry was not supported by the results of this CQI as most clinical outcomes such as intensive care unit admissions, overall mortality and infection-related mortality were not significantly different after implementation of an ASP. The strength of our ASP was that restricted antibiotics could still be empirically prescribed in the initial stage of managing infections. However, doctors had to justify their decisions later to secure a subsequent continued supply. This arrangement ensured the timeliness of initiating life-saving empirical treatments, and at the same time, controlled inappropriate prescriptions. Another strength of our programme was the availability of concurrent feedbacks on the optimal antibiotic selection, dosage and route of administration if the empirical treatment was considered inappropriate. Concurrent feedback ensured timely correction of inappropriate antibiotic prescriptions. This could reduce the amount of inappropriate antibiotics prescriptions, reduced the time lapse before patients received appropriate antibiotics and improved patients’ clinical outcomes. However, concurrent feedback on the same day was not always feasible in clinical setting due to various resource constraints. The availability of dedicated ID experts, pharmacists and input of extra manpower was an important determinant in implementing concurrent feedback on same day.

We observed shortening of hospital stay after implementation of the ASP. However, the causal relationship could not be confidently established from this study, and the reduction in hospitalisation was probably a reflection of difference in case-mix between the two study periods. While some studies have reported a reduction in hospital stay,20 most studies do not support this hypothesis. The overall hospital stay is affected by many potential confounders and it is difficult to determine the impact of ASP precisely. Infection-related hospital stay could better reflect the impact of an ASP and should be considered an important end point in the future research.

Few studies have addressed the correlation between ASP and septicaemia. We observed more septicaemic episodes in the post-ASP period. While this could be explained by difference in case-mix of the admitted patients, the possibility of ASP leading to more septicaemia episodes could not be totally excluded. Whereas this remains a possibility, the causal relationship can be confidently established only after a meticulous review of the whole management process, taking into account the temporal relationships between development of septicaemia, commencement of antibiotics and any subsequent switch of the antibiotics. With the absence of these data, it is beyond the capacity of this CQI programme to establish or refute such a relationship. Future studies should focus on this potential correlation.

Impact of the ASP on antibiotics consumption

Previous studies have shown that ASPs have consistently been effective in reducing prescriptions of restricted antibiotics. Our CQI showed the effect also extended to non-restricted antibiotics. In parallel with the decrement in restricted antibiotics consumption, decrements were observed in some non-restricted classes of antibiotics. The consumption of macrolides and metronidazole/tinidazole was significantly reduced. As the prescription of these classes of antibiotics was not restricted, this favourable observation could be attributed to other components of the ASP, such as intensive education sessions targeting frontline doctors on appropriate choices of empirical therapy for chest infections. Consumption of penicillins increased substantially, which was mainly attributed to increased intravenous co-amoxiclav (Augmentin) consumption. With increasing admissions due to respiratory diseases and chest infections, and with education on correct antibiotics choice, it is likely that doctors have changed their empirical therapy from restricted antibiotics to the penicillins in accordance with the local and international management guidelines on pneumonia.1421

This CQI did not explore the impact of reduced antibiotics prescription on susceptibility pattern of local microorganisms. The 1-year post ASP period might be too short to detect any meaningful change in susceptibility pattern. In addition, the change in hospital infection control practice had to be accounted for in the evaluation. Due to the complexity of inter-related confounders, it would be more rewarding to design a separate study to evaluate the potential correlations.

Financial implication of the ASP

Cost shifting, defined as reduced spending on restricted antibiotics and increased spending on approved antibiotics, was observed in this study.19 Almost all cost reduction was attributed to cost reduction in restricted antibiotics whereas the cost of non-restricted antibiotics, especially the penicillin groups, increased after implementation of the ASP. It was known that antibiotics restriction might shift prescription to more expensive choices or alternatives with similar price. This resulted in cost shifting rather than cost saving. To minimise cost shifting, the choices of restricted antibiotics should be carefully chosen to avoid shifting within the same antibiotics group. An ASP should also focus on reducing total antibiotics consumption by reducing unnecessary initiation of antibiotics or stopping antibiotics when clinical evidence is strong enough to exclude presence of infections.

Some doctors believe that an ASP emphasises cost savings rather than improving quality of patient care. This CQI showed that saving from an ASP accounted for 0.56% of the total hospital bed cost. Hence, it is a fallacy to believe that ASPs only served as an administrative control to curtail hospital expenditure, although it is obvious that through optimal selection and prescription of antibiotics, the inappropriate consumption and expenditure could be reduced. It should be remembered that the ultimate target of an ASP is to improve antimicrobial susceptibility through reduction of unnecessary and inappropriate antibiotics usage.

Our ASP incurred additional human costs, although these were offset by savings from antibiotics expenditure. It was found that doctors’ costs constituted 74.9% of direct ASP costs and many of their responsibilities (such as consultation and feedback) could not be delegated to other colleagues. This observation questioned the sustainability of the ASP as its human cost was high and the cost-effectiveness was expected to reduce as the increment effectiveness (i.e. antibiotic savings) would decrease with time. To ensure sustainable success of an ASP, correct prescribing behaviour should be cultivated among doctors. This can be achieved via regular dispatch of updated information, seminars and education programmes. Through continuous education, timely specialist input, and a smooth administrative process, it is hoped that prescription behaviour can be fundamentally modified towards more healthy practice.


The pre-study and post-study populations were heterogeneous in their demographic characteristics, past medical health, current admission diagnoses, and indications for antibiotics. In general, patients in the pre-ASP period had more chronic comorbidities. However, patients in the post-ASP period were older, and had significantly more respiratory diseases and infectious diseases that would require more antibiotic prescriptions. Outcomes directly related to infection (such as antibiotics consumptions, antibiotics related costs and infection related morbidity) were expected to escalate in the post-ASP period. However, the antibiotic consumption and the related costs improved in the post-ASP period and the infection-related mortality remained unchanged. The impact on other outcomes that were not directly related to infections (such as length of hospitalisation, intensive care unit admission, overall mortality) would be difficult to determine due to the limitation of study design. Prospective controlled studies are required in the future to confirm the causal relationships.

Instead of using a clinical definition of septicaemia, our definition of septicaemia was based on ICD-10 codes. More septicaemia episodes were observed in the post-ASP period. It could be related to difference in case-mix, more blood cultures being taken and inaccurate coding or it could be attributed to the ASP. The observation was purely descriptive and the underlying reason and causal relationship cannot be confidently refuted or confirmed by this retrospective study. Further studies are needed to: describe the prevalence of septicaemia after implementation of an ASP using better defined and accepted clinical criteria; evaluate the impact of an ASP on septicaemia patients; and determine the underlying causal factors.

Not all important clinical end points were captured in this CQI. The time lapse between diagnosis of infection to prescription of appropriate antibiotics, duration of intravenous and oral antibiotics for the individual patient, improvement in appropriateness of antibiotic prescriptions, infection-related hospitalisation rate, infection-related intensive care unit admission, the prevalence of antibiotic-related side effects, acceptance of the programme by clinicians and patients, and impact of the ASP on antimicrobial susceptibility pattern are important outcomes that need to be captured in future studies. Outcomes in certain subgroups of patients (such as patients put on restricted antibiotics only, patients harbouring septicaemia, patients with different underlying comorbidities) were not adequately explored in this CQI programme, but deserved detailed evaluation in future clinical studies.


A multidisciplinary ASP is effective in reducing antibiotic consumption and related expenditure. Our ASP did not adversely affect the quality of patient care as most clinical outcome parameters remained unchanged. Whether an ASP can improve clinical outcomes, antimicrobial susceptibility pattern and quality of life measures requires further evaluation.

Our ASP incurred additional human costs. However, the cost increment could be easily offset by savings from antibiotic consumption. Economic evaluation showed that it was cost-effective to implement such a programme as it reduced both antibiotic consumption and overall costs. It was also technically feasible to implement such a programme as only a relatively small number of essential staff were involved. The generalisability of our ASP to other non-medical departments, non-acute hospitals and outpatient clinics, and the sustainability of such a programme, could be explored further in future studies.


The authors want to express thanks to members of the Antibiotics Stewardship Program Working Group of Queen Elizabeth Hospital (Mr K M Law, Dr W Lam, Dr M Y Chu, Dr S Wong, Dr C Lai) and members of the Hospital Infection Control Team for their time and efforts in running this programme. We would also like to extend our thanks the hospital administration (Dr C T Hung) and members of the Hospital Drug and Therapeutic Committee (Dr C S Li) for their unfailing support and advice.

Appendix A

Breakdown of costs related to the antibiotics stewardship programme (ASP)


  • Competing interests: None.