Pharmacists’ considerations on non-medical switching at the hospital: a systematic review of the economic outcomes of cost-saving therapeutic drug classes

Abstract

Objectives

Non-medical switching (NMS) strategies have the capacity to reduce overall costs in hospitals while maintaining a high level of care. However, the most appropriate diseases and/or medicines for NMS strategies are still vague. The aim of this review was to give a state-of-the-art summary regarding the economic outcomes resulting from the use of NMS strategies and to discuss whether they would be implementable in a hospital inpatient setting.

Methods

A systematic literature search was conducted in Medline, Embase, and ScienceDirect. Studies published between 1988 and 2018 were included if they evaluated the economic impact of NMS strategies or if they performed an economic evaluation between two drugs. Studies regarding antineoplastic agents, endocrine therapies, and immunostimulants, or immunosuppressants, and biosimilars were excluded.

Results

Fifty (69%) studies assessing an NMS strategy and 22 (31%) studies comparing two medicines were allocated to four categories: prospective studies (n=8, 11%); retrospective chart reviews (n=29, 40%); retrospective claims analysis (n=13, 18%); and retrospective data analysis (n=22, 31%). Hypercholesterolemia, peptic ulcer, and gastro-oesophageal reflux diseases, diabetes mellitus, and venous thromboembolism were the most prevalent diseases in studies evaluating an NMS strategy. Sixty-eight per cent of the included papers reported a reduction in costs with no significant changes in health outcomes and 8 per cent reported a deterioration in health outcomes and/or increased costs.

Conclusion

Regardless of the exclusion of studies regarding biologics or medicines used in oncology, the review highlights that NMS strategies with medicines whose management do not require a thorough clinical assessment were associated with reduced costs and no significant changes in patients’ health outcomes, in the inpatient setting. NMS strategies targeting medicines that require an extensive clinical assessment should be evaluated using hospital-specific effectiveness and/or utility data prior to their implementation.

Introduction

Switzerland excels at transforming healthcare spending into health outcomes.¹ However, Switzerland has the second most expensive healthcare system among the member countries of the Organisation for Economic Cooperation and Development.² Healthcare spending reached a new high in 2017 at 83.7 billion (bil.) US dollars ($) (ie, 12.3% of Switzerland’s gross domestic product).^{3 4}

In Switzerland, hospitals were the most expensive supplier of goods and services in 2016 ($28.9 bil., 35.1%) and the most expensive sector of the mandatory health insurance the same year ($13.1 bil., 41%, stationary and ambulatory).^5–8 Medicines were the third most expensive sector of the mandatory health insurance ($6.0 bil., 18%) and represented roughly 5% of Swiss hospitals’ operating budgets.⁹ Medicines have the potential to be a significant source of savings in hospitals.¹⁰

In order to achieve cost savings, all hospitals must rationalise drug use and promote evidence-based medicine.¹¹ Three strategies are widely implemented in hospitals to achieve these purposes: the hospital drug formulary, generic substitution, and non-medical switching (NMS) strategies. NMS strategies allow hospital pharmacists to replace prescribed drugs by chemically different medicines for reasons other than side effects, poor adherence, or lack of clinical efficacy.^{12 13} The substitute medicine may belong to the prescribed drug’s therapeutic class or not, must present similar therapeutic effects, and may be less expensive.^14–16

NMS strategies have the potential to reduce overall costs in hospitals and diminish healthcare costs on a national level while maintaining a high level of care.^16–20 However, there is no explicit recommendations regarding whether an NMS strategy should be established or not. Hence, the most suitable diseases and/or medicines for NMS strategies are still unknown.

The aim of this systematic literature review was to provide a state-of-the-art summary regarding the economic outcomes resulting from the use of NMS strategies and to discuss whether their implementation in a hospital inpatient setting could help lower overall costs while maintaining or improving health outcomes.

Methods

Search strategy

We performed a systematic literature search ranging from January 1988 through June 2018 using the Medline, Embase, and ScienceDirect databases. The Embase and ScienceDirect databases were selected as additional literature sources in order to retrieve papers from drug and pharmacy journals. Twelve medical subject headings terms were selected and combined into 62 associations to build the search strategy (table 1 and online supplemental table 1). The search was limited to papers in French, English, Spanish, and Italian and finalised on 3 July 2018.

Table 1.

Medical subject headings selected to build the search strategy

Medical subject headings
Cost Control	Economics, Pharmaceutical
Cost-Benefit Analysis	Formularies, Hospital
Costs and Cost Analysis	Hospital Costs
Drug Costs	Hospitals, University
Drug Substitution	Pharmacy and Therapeutics Committee
Economics, Hospital	Pharmacy Service, Hospital

Study selection

Study selection followed the study eligibility criteria defined by Participants, Interventions, Comparisons, Outcome(s), and Study design (PICOS) as recommended by the PRISMA statement (table 2).²¹

Table 2.

Study eligibility criteria specified according to PICOS

	Inclusion criteria	Exclusion criteria
Participants	Patients on medication	Individuals with no health condition
Interventions	Non-medical switching	Generic substitution; biosimilar substitution; dosage change
Comparisons	None	None
Outcomes	Economic	No economic outcomes
Study design	Economic evaluation	Reviews; meta-analysis; non-economic evaluations

Studies were included in this review if: they evaluated the economic impact resulting from an NMS between at least two prescription drugs; or if they performed an economic evaluation between potentially switchable drugs. Regarding the definition of the economic evaluations performed in the included studies, we referred to Drummond et al:²² cost analysis (CAs) were deemed as partial economic evaluations because they only considered the compared strategies’ costs without addressing their consequences. Cost-benefit analysis (CBAs), cost-effectiveness analysis (CEAs), and cost-utility analysis (CUAs) were considered as full economic evaluations because they identified and measured both strategies’ input and outcomes in monetary, physical, or quality-adjusted life years (QALYs) units, respectively. Cost-minimisation analysis (CMAs) assumed that the evaluated strategies had an equivalent clinical effectiveness and were also beheld as full economic evaluations.

Study selection consisted of three main screening phases: title, abstract, and full-text screening. Study selection was carried out by two authors independently. When they could not reach an agreement for a particular paper, the third author acted as an arbitrator. Title screening withdrew studies that met no inclusion criteria. The following exclusion categories were applied during abstract screening: A. Biologics substituted with biosimilars. B. Economic evaluations regarding the use or the implementation of a hospital formulary. C. Generic substitution and NMS strategies regarding antineoplastic or immunomodulating agents. D. Guidelines regarding NMS strategies. E. Healthcare economics literature discussing NMS strategies. F. Prescription drugs being substituted with generics. and G. Studies without monetised results. Full-text screening applied the same exclusion categories.

Data abstraction

After a full-text review, we separated the included studies into four categories according to the type of data source used for the studies’ evaluations (table 3). The following information was sought from each study: first author’s last name, year of publication, number of patients and source of patient data, study perspective, evaluated medicines, type of economic evaluation (ie, CA, CEA, CUA, CBA, or CMA), type of sensitivity analysis (ie, deterministic and/or probabilistic), and costs considered in the evaluation (other than wholesale acquisition costs).

Table 3.

Description of the four classification categories

Classification category	Description
Prospective studies	Study patients were followed prospectively before and after the implementation of an NMS strategy.
Retrospective data analysis (RDAs)	NMS data for the analysis was mainly retrieved retrospectively from past trials or peer-reviewed literature.
Retrospective chart reviews (RCRs)	NMS data for the analysis was mainly retrieved retrospectively from medical records, regardless of the provider of healthcare services (ie, physician, pharmacist, hospital).
Retrospective claims analysis (RCAs)	NMS data for the analysis was mainly retrieved retrospectively from claims databases, regardless of the provider of healthcare services (ie, physician, pharmacist, hospital).

NMS, non-medical switch/switching.

Data synthesis

The various NMS strategies were classified according to their Anatomical Therapeutic Chemical (ATC) code.²³ Each drug was assigned to one anatomical main group (AMG) and one therapeutic subgroup (TS). In general, each study evaluated medicines that belonged to the same AMGs and the same TSs. The final number of identified AMGs and TSs was adjusted for studies that evaluated multiple TSs.

Results

Characteristics and outcomes of the included studies

The search strategy yielded 17 170 citations, and no additional records were identified through other sources (figure 1). Seventy-two articles were included in our review and allocated to the predefined categories (online supplemental table 2): eight (11%) prospective studies, 29 (40%) RCRs, 13 (18%) RCAs), and 22 (31%) RDAs. Fifty papers evaluated an NMS strategy and 22 compared two medication strategies without assessing a switch per se (figure 2, online supplemental table 2).

Figure 1.

Flow chart of our review process. See online supplemental table 2 for details regarding the studies included in the qualitative synthesis.

Figure 2.

Details regarding the type of data sources used in the included studies that evaluated a non-medical switch and compared two medications strategies. Total number of studies = 72. Percentages have been rounded and are for information purposes only. Totals do not necessarily add up to 100%.

RCR was the primary type of data source for the evaluation of NMS strategies (n=23, 32%) whereas RDA was the preferred type of data source when the aim was to compare two medication strategies without assessing a switch per se (n=14, 19%). Before RCAs in 2003, RCRs and prospective studies were the primary sources of data for the evaluation of NMS strategies (online supplemental figure 1). In 2018, RCRs were the most frequent sources of data exploited to evaluate an NMS strategy, whereas prospective studies were gradually abandoned.

Over two-thirds (68%) of the included papers that evaluated an NMS strategy reported a reduction in costs without a significant change in health outcomes. Only six studies reported a deterioration in health outcomes and/or increased costs (online supplemental table 3).

Eighty-two per cent (n=41) of the studies performed a partial economic evaluation (ie, cost analysis). In contrast, the most used form of full economic evaluation was CEA with six studies (online supplemental table 4).

Regarding costs and sensitivity analysis, about two-thirds (62%) of the studies that evaluated an NMS strategy considered resource utilisation costs (eg, labour costs, hospitalisation costs, emergency room visits, and so on) in addition to wholesale acquisition costs (online supplemental table 5). A higher proportion of studies (73%) did not perform any kind of sensitivity analysis (ie, deterministic sensitivity analysis or probabilistic sensitivity analysis) to straighten the robustness of their results (online supplemental table 6).

Most prevalent anatomical main groups and therapeutic subgroups for non-medical switching strategies

Based on the ATC classification system²³ seven AMGs were identified in the studies evaluating an NMS strategy. AMGs such as cardiovascular system, alimentary tract and metabolism, or blood and blood-forming organs were the most represented in this review with 19 (37%), 13 (25%), and seven (14%) studies, respectively (online supplemental table 7). Almost every AMG was evaluated at least once with studies using RCR as a data source (online supplemental figure 2).

Taken altogether, half of the studies that belong to the cardiovascular system evaluated antihyperlipidemic agents (ie, HMG-CoA reductase inhibitors [n=9, 17%]). In alimentary tract and metabolism, seven papers (13%) studied peptic ulcer and gastro-oesophageal reflux diseases and five studies (10%) covered diabetes mellitus. Blood and blood-forming organs included four studies (8%) about antithrombotic agents and three (6%) studies regarding antianemic preparations (online supplemental table 7).

Discussion

Easily implementable non-medical switchings in a hospital inpatient setting

Statins

NMS of HMG-CoA reductase inhibitors (statins) should be an opportunity to reassess the patient's cholesterol-lowering treatment in order to improve their health outcomes. These NMS strategies may reduce overall costs in hospitals’ inpatient settings and should be examined locally by means of full economic evaluations. Among the nine studies that evaluated statins in our review, two papers^{24 25} stand out. The first paper²⁴ considered resource utilisation costs (RUC) in addition to whole acquisition costs (WAC) and concluded that an NMS from pravastatin to simvastatin increased total costs but improved the lipid control of the switched patients. The other one²⁵ considered a broader panel of RUC and concluded that an NMS from atorvastatin, fluvastatin, or pravastatin to either cerivastatin or simvastatin not only improved lipid control in switched patients but also reduced costs. Furthermore, none of the seven other studies found lipid controls to be negatively impacted.

The collaboration between the key stakeholders surrounding the patient (ie, the pharmacy and therapeutic committee, the physicians, the pharmacists, and the nursing staff) will ensure safe NMS from a statin to another.²⁶ Combined with the opportunity to avoid the exorbitant costs of later coronary heart disease complications and statins’ interchangeability at comparable dose, each institution should consider statins’ NMS to improve patients' health outcomes.^26–31 Hence, even if NMS among statins do not guarantee immediate cost reductions, later savings are to be expected thanks to the improvement of the patients’ health outcomes.

Peptic ulcer and gastroesophageal reflux diseases

Based on WAC alone, the NMS of medicines used for the prevention or the treatment of peptic ulcer and gastro-oesophageal reflux diseases (ie, H₂-receptor antagonists [H2RAs] and proton pump inhibitors [PPIs])) may reduce overall costs without any impact on the patient's health outcomes. Among the seven studies in our review, three cost analyses evaluating the NMS of H2RAs concluded that their NMS reduced overall costs.^32–34 Good et al,³³ bore the most robust results because the study included RUC and the largest number of patients (704). Four papers^35–38 evaluated the NMS of PPIs but diverged in conclusions: three of them^{35 37 38} only considered WAC, performed no sensitivity analysis, and concluded that the studied NMS strategies reduced overall costs: the last one³⁶ stated that an NMS from omeprazole, lansoprazole, or pantoprazole to either omeprazole or rabeprazole was associated with higher costs and poorer health outcomes.

Gaebel et al,³⁶ highlighted a resumption of the patient's gastrointestinal symptoms as well as new symptoms following the NMS. This apparent loss of efficacy and emerging of adverse events led to an additional and unanticipated healthcare service use (ie, ambulatory office visits and hospitalisations) and exceeded the forecasted savings based on WAC alone. However, because of the small sample size (40 patients), the outpatient setting, and the unique population of the study (ie, aboriginals living in remote and isolated areas of the Northwest territories of Canada), we think that their results should not invalidate the findings of the other studies. Instead, it should challenge the implementation of NMS strategies in the outpatient setting.

According to our review, the failure of proper patient follow-up seemed to be the main reason for NMS strategies to increase overall costs and/or deteriorate patients’ health outcomes. It was highlighted in four of the six studies whose NMS strategies resulted in a deterioration of the patients’ health outcomes and/or increased costs.^{36 39–41} Eventually, increased RUC resulting from hospitalisations, emergency room visits, and management of new adverse events exceeded the expected savings in WAC.

Intricate non-medical switchings in a hospital inpatient setting

The rest of the literature in our review concerned diseases whose management required a extensive clinical assessment. Hence, these diseases’ medicines were used for specific clinical events rather than for general indications, as opposed to statins or PPIs. Thus, it was too difficult to estimate if the NMS of such medicines should be implemented in a hospital inpatient setting.

We will briefly discuss diabetes mellitus and venous thromboembolism (VTE) as they were the most prevalent diseases in the rest of our review’s literature.

Diabetes mellitus

Economic models based on Markov cycles may be the best way to assess medications' strategies surrounding diabetes mellitus. Because of the complexity and the nature of diabetes mellitus, drug strategies surrounding this disease should be evaluated using full economic evaluations with a lifetime horizon. Interestingly, three studies^42–44 used a validated computer simulation model to assess the NMS of insulins in South Korea, Thailand, and China, respectively. (This model was designed to convert short-term clinical effects into long-term economic outcomes and has been described elsewhere.)^{45 46} Hence, using RUC in addition to WAC, effectiveness data, and utilities values, they were able to calculate incremental cost-effectiveness ratios and QALYs whose robustness were challenged using either deterministic (DSA) or probabilistic (PSA) sensitivity analysis. Nevertheless, their conclusions only apply to their very specific population and might be hardly generalisable.

Such economic models could be able to tackle issues that would otherwise be difficult to overcome in standard studies because of time or costs constraints. For example, another study,²⁵ that evaluated the same NMS strategy through an RCR, suffered from a small sample size, the absence of a control group, a limited time horizon, and only considered WAC. Depending on the availability of the data, such limitations may be overcome with the help of an appropriate economic model.

Venous thromboembolism

As with diabetes mellitus, our review could not safely indicate if the NMS of medicines used for VTE’s prevention should be implemented in a hospital inpatient setting. Each included study was from a different perspective, considered an incomparable number of patients with different diagnostics, and evaluated unique NMS strategies that involved different therapeutic subgroups, making comparisons extremely difficult. Considering the delicate dosing required by the medicines used in VTE’s prevention, the consequences associated with underdosing (ie, deep venous thrombosis and pulmonary embolism) or overdosing (ie, haemorrhages) and the myriad of clinical conditions that could lead to VTE: mere cost analysis, as the ones included in our review, were not sufficient to faithfully evaluate the NMS of these medicines. Hence, such complex diseases could also be better evaluated using an economic model.

Limitations

Several limitations are worth mentioning in our review. First, we did not look for additional papers in grey literature databases (eg, trial registries, conference abstracts, and so on) nor did we snowball additional results from the included studies. Grey literature databases might have contained pertinent papers to refine our conclusions. Nevertheless, such unpublished papers did not undergo any rigorous peer-review process and might have been detrimental to our review as well because of their potential unidentified bias.

Next, the therapeutic subgroups identified in our review represented only a fraction of the drug budget in hospitals. Even though hospitals' budgets are dominated by antineoplastic agents, endocrine therapies, and immunostimulants or immunosuppressants, we excluded such economic evaluations because we believe that such medicines should be reviewed independently, given their highly specific protocols. Also, we think that results regarding less complex or critical treatments such as statins, H₂-receptor antagonists, or PPIs might be more easily generalised to other settings. Furthermore, even though such medicines might not generate colossal savings after the implementation of an NMS strategy, they might still improve patients’ health outcomes in the long term at the price of an adequate follow-up and deserved to be highlighted.

Finally, we did not perform any quality assessment regarding the included studies. The use of a scale for assessing the quality of the studies could have strengthened or qualified our conclusions, even if these scales come with their own limitations. This tool also could have been used to categorise the included studies differently. Hence, no weights or scoring methods were considered within the four categories. Instead, we used the studies’ perspective, number of patients, type of costs included, and type of sensitivity analysis performed to appraise their quality.

Conclusion

NMS strategies found in the literature were tested in various settings and with a multitude of medicines. Regarding hospitals’ inpatient setting, the NMS of medicines that do not require a thorough clinical assessment can be safely achieved. Even though these medicines are no longer the major cost-drivers of the hospitals budgets, NMS of these drugs has the potential to improve patients’ health outcomes, rationalise hospitals' therapeutic array, and can be easily implemented. The evaluation of the NMS of medicines used for the management of more intricate diseases is much more difficult and require additional studies. Furthermore, the results from such NMS strategies may be less or non-generalisable as they require population-specific effectiveness and/or utility data. Ideally, they should be evaluated prior to their implementation using hospitals’ medical chart data.

Whether easily or hardly implementable, NMS strategies should be considered only if economic promises ensure equivalent or improved patients’ health outcomes.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.

Ethics statements

Patient consent for publication

Not required.