Health economic studies of antimicrobial stewardship programmes: A scoping review

Patient and Public Involvement

Patient partners were not involved in the design of this scoping review.

Search strategy

We searched for health economic studies of antimicrobial stewardship programmes using the databases PubMed, EconLit, Business Source Alumni Edition, Business Source Premier and CINAHL Pluss. Our search strings included the following terms: ((cost-effectiveness) OR (cost-benefit) OR (cost-utility)) AND ((antimicrobial stewardship) OR (antibiotic stewardship)) AND ((hospital) OR (secondary care)). Search and selection of studies was carried out between April and May 2022 and October 2023 (in PubMed). We also conducted a hand search of studies by screening the references of studies retrieved through the above research strings.

Eligibility criteria

We limited our review to studies conducted within a hospital setting and published in English between January 2002 and October 2023. We selected studies measuring both costs and clinical effectiveness of the intervention. Cost measures had to include the intervention cost and could include in addition, antibiotic costs or hospital costs. We selected only studies that included a valued (i.e. clinical) outcome (death, length of stay, Clostridioides difficile infection), excluding studies that only mention process (mediating) outcomes such as DDD and total antibiotic use. We excluded antifungal stewardship programmes and studies using diagnostic test innovations as a mechanism to improve antibiotic utilisation (e.g., procalcitonin tests and rapid diagnosis tests).

Study selection

The titles and abstracts of studies returned by our database were searched to select potentially eligible studies. We retrieved and scrutinised those studies to make our final selection. Full texts were scanned by LA and PN.

Data extraction

Papers would be categorised into three groups based on the method used: cost-effectiveness or cost benefit. To categorise papers based on these three different methods, we used the following definitions of economic evaluations, according to the definitions used by Shiell et al.¹⁵ as follows:

No costs were reported in UK pounds sterling. To enable consistency and fair comparisons between studies, we firstly converted costs to pounds sterling using the exchange rate at 30 June in the year of costing (https://www.xe.com/currencytables/). Secondly, costs in GBP were inflated to 2021/22 values using the NHSCII pay & prices index; the 2022/23 index was not available at the time of writing, so the 2021/22 indices were re-used to deflate costs obtained at 2022/23 prices (https://www.pssru.ac.uk/unitcostsreport/)¹⁶. For costs in 2018, we started at the 2018/19 NHSCII value, for example. We report base-case results as provided by the authors.

Articles reviewed

Two hundred and twenty-eight articles were returned using the search strings described above and the further search for eligible studies identified amongst the references returned seven articles to give a total of 237 studies (Figure 1).

Figure 1. PRISMA diagram of the selection of economic evaluation studies of ASP.

One hundred and ninety of the two hundred and thirty-seven studies were rejected after examination of title and abstract, leaving forty-four studies for full text assessment. Note, that in line with our stated aims, we only included studies that provided intervention costs and a summary measure of effectiveness or utility/benefit. This means that we did not include studies such as Mahmoudi et al.¹⁷ and Alawi et al.¹⁸ that did not specify the costs of the intervention. Eight studies (Table 1 and Table 2) were then selected after reading the full articles^19–26. Outcomes included CDI, length of stay (LOS) and mortality. Only one study was based on a Randomized Control Trial (RCT)²¹. The other studies were observational studies – one of these based on secondary data - and all took a health service perspective rather than a broader societal perspective. Thus, none of the studies attempted to capture the effects of payback that might occur to society more broadly as a result of lower level of resistant organisms in the environment. We return to the issue of potential societal types of paybacks in the discussion.

Types of study

We found six studies that fulfilled our criteria in the cost-effectiveness category and two in the cost-utility category. We found no cost-benefit studies as defined above (although we did find a cost-effectiveness study (Slayton et al.²²) that was described as a cost-benefit study by the authors).

Cost-effectiveness studies

The six studies that included the intervention cost and at least one clinical effect are described in Table 1. The table includes the single randomised trial recorded in our review²¹. This was a step-wedged cluster trial conducted in nine hospitals. There were four observational studies and one purely literature-based study²⁰. The intervention dominated conventional care in two of the six studies^22,23 i.e., it reduced overall costs net of the intervention cost and improved health outcomes. Four studies used number of infections as an outcome measure, but not consistently: for example, Slayton et al.²² only considered CDI while Lin et al.²³ used all healthcare-associated infections (HAIs).

Cost utility studies

The two cost-utility studies reported here are both associated with the use of ASP for treatment of bloodstream infections (Table 2). Types of costs included in the analyses were hospital costs, blood culture costs, costs of implementing the computerized clinical decision system (CCDS) and staff time costs. The effectiveness parameter was measured as quality-adjusted life years (QALYs) in both studies. The intervention dominates the control in Gebretekle et al.²⁵, with a saving of £1,700 per QALY gained. The intervention reported by Scheetz et al.²⁶ was highly likely to be cost-effective, with a cost per QALY gained of £1,500.

Literature

Our scoping review found six studies on the cost-effectiveness of ASPs and two cost-utility analyses. Many other papers included details of medication cost savings resulting from ASP implementation but did not measure clinical outcomes and/or did not provide details of intervention costs.

Cost estimations

Hospitalisation costs in the form of bed days have a large impact on total costs and were used as an/the outcome in four of the studies^19,22,25,26. Overall, the studies provided mixed results on the effect of ASPs on length of stay/readmission. This is mirrored by the broader literature. For example, while Hagert et al.²⁷ did not find any difference in length of stay for patients in the ASP compared to the control group, Niwa et al.⁷ reported a reduction of one day in the average length of hospital stay following implementation of an ASP. A reduction in Clostridioides difficile infection rates, as a result of effective ASPs, will reduce hospitalisation costs and is likely to have major impact on the cost-effectiveness analysis. Many clinical studies have investigated the effects of CDI on hospital stay and results show that CDI is associated with longer length of stay and consequently is an important driver for hospital costs²⁸. Previous analysis of the effects of CDI on healthcare costs have shown that direct healthcare expenditures and opportunity costs associated with additional LOS were the main CDI cost drivers²⁹. One study included in this review specifically estimated the cost savings associated with the reduction of CDI rates²².

Only one study captured the effects of early switch from intravenous to oral administration²¹. It is worth noting that many authors have investigated the effects of early switching antibiotics from IV to parental administration, showing either positive outcomes for patients or decreasing healthcare costs^30–32.

Implementation of ASP had cost implications for those in the workforce who are involved at the development, implementation, and operational levels. Two studies provided detailed information about the different types of health staff associated with the ASP intervention^21,22. Van Daalen et al.²¹, for instance, reported staff costs in detail, including activities of coordinator and local supervisor in planning and sharing materials with the team, giving briefings and travelling, along with physicians’ workloads (such as time specialists and residents spend attending briefings or performing e-learning). Workload of medical doctors in charge of implementing ASP is an important issue since increasing their responsibilities associated with the new intervention may affect the capacity of physicians to better prescribe, as showed in qualitative study conducted in India by Baubie et al.³³ Other studies provide scant detail on costs (e.g. Lin et al.²³), such that the costs provided are likely to be an underestimate of total costs, meaning that cost-effectiveness will be overestimated.

Effectiveness measures

Several effectiveness outcomes were used in the economic evaluation studies retrieved for this scoping review. Two studies used mortality rates as an effectiveness measure^19,20. One cost-effectiveness study used 30-day mortality rate as the main effectiveness outcome and DDD and resistant infection rates as secondary outcomes¹⁹. Another study used CDI rates and potential additional deaths associated with CDI to estimate the cost per avoided resistance and cost-per-life-year gained (LYG)²⁰. The use of mortality rate to assess the quality of hospital care has been discussed in more detail elsewhere by Lilford and Pronovost³⁴ indicating that standardised mortality rates may not be reliable to judge the performance of healthcare services.

One study reported a reduction in HAI rates but an increase in readmission rates (although neither effect was statistically significant), alongside an overall cost saving, driven by a reduction in drug and antibiotic costs²⁴. Without converting the clinical effects to QALYs in a cost-utility analysis, it is not possible to determine whether the intervention has an overall positive or negative impact on health, or whether increasing readmissions is justified for achieving cost-savings.

The studies included here report on antimicrobial consumption and the clinical benefits that may flow from reduced iatrogenic infections, especially Clostridioides difficile. However, none of the interventions specifically considered the potential effect of ASPs in better targeting therapies as a result of Bug-Antibiotic Mismatch (BAM). ASPs have the potential to reduce mismatch by prompting clinicians to change the antibiotic prescription in response to results from the microbiological laboratory recorded in the electronic record.

Lastly, the conclusions of Slayton et al.²² are likely to be optimistic when applied to an ASP alone since it was a multi-component intervention not limited to ASP.

Societal perspective

As stated in the Introduction, society as a whole benefits from improved antibiotics prescribing. However, none of the studies captured the potential benefits and cost (savings) of reduced drug-resistant infections resulting from a reduced micro-biological resistance in the environment. We speculate that this is not because scholars have not considered this issue, but because of methodological difficulties. One barrier to such studies is the difficulty of estimating the contribution of improved prescribing in any domain (community or hospitals) to the overall burden of resistant organisms in the environment. It is also difficult to capture the burden of antibiotic prescribing on other hospitals, other health care settings and in veterinary and agricultural practice. We have identified a research group that is tackling this challenging problem³⁵.

Economic models

The included studies use different methods for the economic analysis and collect data from distinct sources. Furthermore, they cover different clinical scenarios, very different intervention (ASP) types and different types of institution in different places. Faced with such massive heterogeneity of method, topic and context, and a limited number of studies, it would be inappropriate to draw general conclusions. Rather the papers are instructive of the many methodological challenges such as the choice of effectiveness parameters in analysing a service delivery intervention such as ASPs³⁶.

Implications

The extreme heterogeneity of studies included in this review has a number of implications for future research. First, there is a strong argument for standardisation of outcomes (costs and effects) across methods in terms of costs and other outcomes included. We recommend that an international meeting be convened to see if such agreement can be forged. Second, it is unlikely to be sensible to combine effectiveness parameters for ASP interventions in a quantitative meta-analysis. It is difficult to see how one summary measure can inform a future decision notwithstanding random effects methods. If there were very large numbers of individual studies then it might be possible to subgroup by, for example, intervention type or to construct more complex multi-level meta-regression³⁷. Absent from such a large ensemble of studies, we do not think summary measures of effectiveness across studies are useful. Third, pending the evolution of a very large number of individual studies, we advocate the development of an agreed model relating intervention to outcomes. The individual decision makers could then populate such models with parameters estimates informed by literature but adapted to local circumstances. Such a model would also inform the mediating (process) and outcome variables that should be collated routinely.

Limitations

A limitation of this scoping review is that only eight studies were identified for inclusion. Hence, any attempt to produce a summary measure of ASP interventions is limited by the paucity of studies, and heterogeneity of intervention types, variation in the implementation contexts and the different methodological approaches.