This systematic review followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) guideline. A systematic search was performed across 3 major electronic databases (PubMed, Scopus, and EBSCOhost) to identify studies investigating economic evaluations of DHT to improve medication safety outcomes (ie, reduction in ADEs and medication errors). The search strategies included 3 categories of terms related to:

Full details of the search strategy are provided in Table 1. An additional search of reference lists of relevant reviews and included studies was performed.

We included studies that reported health economic evaluations of DHT to improve medication safety outcomes (ie, ADEs and medication errors) compared to standard care. Detailed information related to inclusion criteria is provided in Table 2. We excluded studies where the intervention did not include DHT, studies that did not report medication safety outcomes, studies without full economic evaluation, non-English studies, conference abstracts, and editorials.

The database-screening results were exported to the Mendeley reference manager library and examined for duplicates. Two investigators (authors WNI and NZ) independently conducted a full-text review after screening the preliminary titles and abstracts. Any discrepancies between the 2 reviewers were resolved through discussion. We adopted recommendations from the Panel on Cost-Effectiveness in Health and Medicine for data extraction. Data extracted included (1) study characteristics; (2) clinical and economic outcomes, including the incremental cost-effectiveness ratio (ICER), the cost-benefit ratio (CBR), and the return on investment (ROI); (3) cost components; (4) methodological challenges; and (5) quality of reporting. All monetary values were converted to 2024 US dollar values using the Campbell and Cochrane Economics Methods Group–Evidence for Policy & Practice Information Centre Cost Converter [27].

Reporting quality was assessed using the CHEERS (Consolidated Health Economic Evaluation Reporting Standards) 2022 checklist. The checklist included 28 items, classified into 6 categories: (1) title and abstract, (2) introduction, (3) methods (including choice of model, health outcomes, and measurement of effectiveness), (4) results, (5) discussion, and (6) others. Based on the reporting quality, the included studies were categorized as excellent (score 100%), good (score 75%-99%), moderate (score 50%-74%), and low (score ≤49%) [28].

A total of 408 citations were retrieved from the electronic databases. After removing duplicates, 355 (87%) papers remained for evaluation based on titles and abstracts, yielding 49 (13.8%) records eligible for full-text assessment. A total of 13 (26.5%) economic evaluation studies [10,12,15,17,29-37] were finally included in this systematic review. Most of the studies included CEA (n=9, 69.2%) [12,15,17,29-34], followed by CBA (n=3, 23.1%) [10,35,36], and CUA (n=1, 7.7%) [37]. The included studies were predominantly conducted in hospital inpatient settings (n=10, 76.9%) [10,12,17,30-36]. The study selection flowchart based on PRISMA guidelines is provided in Figure 1.

The most frequently implemented DHT intervention was the CDSS/CPOE or electronic medication record system. This intervention typically replaced traditional paper-based prescribing, with a safety alerts feature to assist clinicians in making informed decisions during the prescribing process [12,17,29,30,33-35]. A wide range of features were used in the included studies, ranging from basic DDI to the addition of medication administration tracking and various alerts, including pregnancy contraindications, allergy checks, dosage checks, therapeutic duplications, and potentially inappropriate medication (PIM) usage in vulnerable populations at increased risk of ADEs [12,17,29,30,33-35]. One study [30] compared a structured pharmacist medication review supported by a CDSS with a review conducted without CDSS support for older hospitalized patients. The intervention included medication reconciliation and personalized pharmaceutical care (Table 3).

Of the 13 studies, 4 (30.8%) [10,31,32,36] showed that the implementation of automated medication-dispensing systems has been increasingly used to reduce dispensing and administration errors, enhance workflow efficiency, and improve stock inventory tracking. These systems typically incorporated automated individual unit-dose dispensing, which ensured precise medication packaging tailored to patient-specific prescriptions, unlike manual systems where nurses prepare doses manually [31,32,36]. The systems were often integrated with barcode scanning technology, either at the hospital pharmacy or at the patient bedside level, enabling verification of both the medication and the patient’s identity to prevent administration errors [31,32]. Additional functionalities included structured drug storage and controlled access (eg, automated dispensing cabinets [ADCs] and medicine carousel systems with rotating shelves or bins to minimize the time spent searching for medications and medication selection errors) [10,32,36]. One intervention also improved inventory management by providing automated real-time stock monitoring [10]. Furthermore, some systems were linked with electronic medication administration records and other CDSS tools, allowing better coordination between dispensing and prescribing processes [31,32,36].

A technology-based pharmacist outreach was implemented in 2 (15.4%) studies [15,37], involving pharmacists engaging directly with other health care professionals to target specific high-risk prescribing errors, such as prescribing nonsteroidal anti-inflammatory drugs (NSAIDs) without proton pump inhibitors (PPIs) for patients with a history of peptic ulcers, beta-blockers for patients with asthma, and angiotensin-converting enzyme (ACE) inhibitors or diuretics without proper monitoring of renal function and electrolytes. The interventions involved a CDSS-supported feedback system and pharmacist educational outreach for general practice staff. This intervention model extended beyond simple error reporting by providing pharmacist-intensive support and guidance to other health care professionals.

The DHT interventions were effective in reducing ADEs, with an average reported effectiveness of 37.12% (range 8.2%-66.5%) across the included studies (Multimedia Appendix 2 and Table 4) [12,17,29,30,34,35]. Medication errors reduced by an average of 54.38% (range 24%-83%) [10,12,15,17,29,31-33,36]. The variability in the effectiveness of DHT in reducing ADEs and medication errors can be influenced by differences in DHT features and the target population. A comprehensive CDSS/CPOE with both prescription entry and administration tracking resulted in greater reduction in ADEs compared to a CDSS/CPOE with only basic alerts [12,34]. Nevertheless, none of the studies provided data on alert compliance tracking by health care professionals. Higher effectiveness was also observed in DHT interventions targeting high-risk populations, such as older and pediatric hospitalized patients. These populations are more prone to ADEs and medication errors resulting from comorbidities, concomitant medications, and differences in physiological characteristics. Consequently, interventions targeting these populations showed greater absolute reductions in ADEs and medication errors compared to interventions targeting lower-risk populations, as the elevated baseline risk provides more room for improvement [29,30].

Of the 13 studies 6 (46.2%) [10,12,17,31,32,35] used a quasi-experimental design (ie, before-after and interrupted-time-series approaches, with multiple data points tracking ADEs and medication errors before and after DHT initiation). A decision analytic model was used in 3 (23.1%) studies [29,33,35], focusing on short-term, event-specific analysis, while a Markov model was used in 1 (7.7%) study [37]. Randomized controlled trial (RCT)–based evaluation was used in 2 (15.4%) studies [15,30]. Challenges related to methodological aspects of evaluating DHT interventions for medication safety are presented in Table 6. Methodological challenges included short follow-up times, the absence of CDSS alert compliance tracking, a lack of ADE severity classification, and the omission of indirect costs (eg, productivity loss and caregiver time).

Based on the quality assessment using the CHEERS checklist, most of the included studies (n=10, 76.92%) [12,15,17,29-33,36,37] were rated as good, and the remaining 3 (23.1%) studies [10,34,35] were rated as moderate. Adherence to the checklist items varied across the section. Most studies did not report any approach to engagement with patients in the study design and results, except for the RCT-based study [30]. The rationale for selecting the model was inadequately reported in over a third of the studies, indicating issues in validating the selected methodology [10,30,31,34,36]. All studies reported how to measure ADEs and medication errors; however, the valuation of these safety outcomes was not reported in some studies [10,12,33]. The components of the outcomes included costs related to additional treatments and hospital stays due to the occurrence of ADEs and medication errors. Although most studies conducted sensitivity analysis to assess the robustness of the findings, only 1 (7.7%) study [32] assessed the heterogeneity of the outcomes based on different types of automated medication-dispensing systems. Most studies only focused on aggregated data without exploring subgroup differences, such as variations within patient populations, hospital wards, or intervention types.

This is the first systematic review that assessed the economic evaluations of DHT interventions to improve medication safety outcomes. More than half of the studies (n=7, 53.9%) focused on a CDSS/CPOE, less than a third (n=4, 30.8%) on automated medication-dispensing systems, and the remaining (n=2, 15.4%) on pharmacist-led outreach programs targeting health care professionals. On average, DHT interventions reduced ADEs by 37.12% and medication errors by 54.38%. In 92.3% (12/13) of the included studies, the DHT was either cost-effective or cost beneficial compared to standard care. Despite a significant upfront cost, DHT showed an ROI within 3-4.25 years. Key methodological challenges included short follow-up periods, a lack of ADE severity categorization, the absence of alert compliance tracking, and the omission of indirect costs.

A CDSS/CPOE has been increasingly used to support clinicians in making informed decisions by providing recommendations based on patient data and clinical guidelines [39]. Nevertheless, the effectiveness of a CDSS is often compromised by alert fatigue, which occurs when clinicians override a large number of potentially irrelevant drug safety alerts (eg, drug interactions that are not clinically significant, flagging dosages outside the standard guideline when the prescribed dosage is appropriate for the patient condition) [40-42]. A previous study showed that alert fatigue might be reduced by using a CDSS/CPOE with interactive features, such as incorporating tiered safety alerts with varying priority levels, offering action plans for clinicians (eg, dose reduction), and requiring clinicians to justify overriding an alert [41]. In our review, none of the studies tracked CDSS/CPOE alert compliance, which may influence their effectiveness. A previous study showed drug allergy alerts had the highest compliance [43]. Further DHT evaluation studies should investigate how alert compliance affects the outcomes to ensure optimal utility of DHT.

In addition, tailoring clinical roles within a CDSS can also enhance alert acceptance [44-46]. A pharmacist-mediated CDSS has improved prescriber acceptance by filtering irrelevant advice and providing actionable drug safety recommendations [44]. This approach leverages pharmacists’ expertise related to medication to support the decision-making process and reduce alert fatigue among prescribers [47-49]. Several included studies combined a CDSS with pharmacist-led interventions, such as structured medication review and targeted training for health care professionals for error correction, highlighting the central role of pharmacists in medication safety [15,30,37].

Advancements in CDSSs/CPOE have significantly reduced prescribing errors, but administration errors, such as administering the wrong dose, using the incorrect route of administration, and failing to administer a scheduled dose, still present substantial room for improvement [50]. This review found that a CDSS/CPOE with integrated prescription entry and administration tracking that address prescribing errors, while also monitoring and mitigating administration errors, achieves a greater reduction in ADEs compared to systems with basic features [12,34]. The development and implementation of more comprehensive systems that target all stages of medication management, including administration, are essential to reduce medication errors and ADEs.

Different types of automated medication-dispensing systems were used in the included studies. Several factors need to be considered when selecting an automated medication-dispensing system, including the volume of prescriptions and types of medication the device can handle (eg, oral tablets, liquid medications, injectables), integration with the existing information technology system, and the presence of an error-checking mechanism [51,52]. Almaki et al [53] and Tsao et al [54] showed that integration with existing digital infrastructure is key to supporting a seamless workflow process to ensure all components of medication management (ie, prescribing, dispensing, and inventory control) are interconnected, enhancing efficiency and medication error prevention.

All the included studies focused on hospital-based automated medication-dispensing systems, with limited investigation in outpatient settings [10,31,32,36]. Williams et al [52] demonstrated that automated medication-dispensing systems for chronic medication regimens, such as antiretroviral therapy, enable efficient one-time password (OTP)–based medication collection and reduced waiting times, benefiting high-volume, resource-limited outpatient clinics. Further studies may adopt this system for less complex medication regimens to improve patient satisfaction and reduce the staff burden, enabling health care providers to focus on other critical tasks, such as optimization of drug therapy [54,55].

Targeting medication errors with a substantial clinical impact was a key strategy in the technology-based pharmacist-led outreach [15,37]. As around 1% of medication errors lead to actual harm, prioritizing interventions is essential [24]. The National Coordinating Council for Medication Error Reporting and Prevention (NCC MERP) developed an index to standardize the classification of medication errors based on severity (ie, no error, error-no harm, error-harm, and error-death), which has been adopted in the hospital setting [56,57]. An initiative to standardize the interception of medication errors using this severity classification index has proven effective to prevent clinically significant patient harm and provide substantial cost savings to the health system [58].

Various economic models were used in the included studies (eg, quasi-experimental design, RCT, decision analytic model), indicating the complexity and multifaceted nature of DHT for medication safety. Sculpher et al [59] emphasized the importance of placing an RCT within a broader framework of evidence synthesis and decision analysis in economic evaluation studies to balance internal validity and broader applicability. Quasi-experimental design can be practical for assessing real-world effectiveness and scalability. However, the assessment of confounders that might affect the observed trend and the lack of guidance in conducting economic evaluations alongside quasi-experimental trials necessitate careful interpretation of the findings [60].

This study is the first systematic review investigating economic evaluations of DHT interventions to improve medication safety outcomes. We conducted a rigorous literature search with a comprehensive strategy encompassing a wide range of DHT intervention types to provide a thorough review of different strategies to improve medication safety. We also included evaluations on methodological challenges in DHT intervention assessment to inform future research direction.

Nevertheless, our review has several potential limitations. First, the generalizability of findings may be restricted to high-income settings, as most included studies were conducted in such contexts. Second, heterogeneity in health care settings, study designs, and economic evaluation methods may hinder direct comparisons across studies. Despite this, a narrative synthesis was developed to integrate and interpret the findings. Third, the exclusion of non-English studies may have limited the comprehensiveness of the evidence base. Finally, there was a lack of medication error and ADE severity classification among the included studies. Since different medication errors and ADEs have varying cost implications, this variability makes the interpretation of overall cost-effectiveness less straightforward.

DHT interventions are economically viable to improve medication safety, with substantial reduction in ADEs and medication errors. On average, DHT interventions reduced ADEs by 37.12% and medication errors by 54.38%. In 92.3% of the included studies, DHT was either cost-effective or cost beneficial compared to standard care. Despite a significant upfront cost, DHT showed an ROI within 3-4.25 years. The key drivers of cost-effectiveness include reductions in outcomes, the proportion of errors resulting in ADEs, and implementation costs. Key methodological challenges included short follow-up periods, the absence of compliance tracking, the lack of ADE severity categorization, and the omission of indirect costs. Future studies should prioritize incorporating alert compliance tracking, ADE and medication error severity classification, and the evaluation of indirect costs, thereby increasing clinical benefits and economic viability.