Automated dispensing cabinets and nurse-related medication errors in inpatient settings: A systematic review

Abstract

Background

Automated Dispensing Cabinets (ADCs) are computerized medication storage and dispensing systems increasingly used in inpatient hospital settings to support safe medication administration. As nurses represent the final point of medication delivery to patients, understanding the impact of ADCs on nurse-related medication errors is critical. This systematic review aimed to assess the association between ADC implementation and medication errors involving nurses in inpatient hospital settings.

Methods

A systematic search of PubMed, MEDLINE, Scopus, and Google Scholar identified 16 eligible studies published between 2008 and 2025 from 184 retrieved records. Included studies employed quantitative, qualitative, and mixed-methods designs. Risk of bias was assessed using the Mixed Methods Appraisal Tool (MMAT), and findings were synthesized narratively due to heterogeneity in study designs and outcome measures.

Results

Across quantitative studies, ADC implementation was generally associated with reductions in reported medication errors, particularly omission errors and wrong drug or wrong dose events, although reported reductions varied widely. Some studies reported maintenance of previously low error rates. Qualitative findings described perceived improvements in workflow efficiency, medication access, and nurses' confidence in medication safety, alongside temporary workflow challenges during early implementation.

Conclusions

Overall, the evidence suggests that ADCs may support safer medication administration and nursing workflows when appropriately implemented. However, evidence quality is limited by largely observational designs, heterogeneous error definitions, and inconsistent outcome reporting, highlighting the need for more standardized and methodologically robust research.

Highlights

• •ADCs reduce key nurse-related medication errors in inpatient settings.
• •Review synthesizes findings from sixteen international hospital studies.
• •Omitted and wrong-drug errors show the most consistent improvement.
• •Nurses report higher workflow efficiency and safer medication access.
• •Study identifies evidence gaps and priorities for future ADC research.

1. Introduction

Medication errors are preventable events that may cause or lead to inappropriate medication use or patient harm, and they are recognized globally as a major contributor to avoidable morbidity and mortality.^1,2 These errors can occur at any stage of the medication-use process, including when prescribing, transcribing, dispensing, and administrating medications.³ Medication administration is a core component of inpatient care and represents the final point at which harm can be prevented before a medication reaches the patient, making administration errors particularly consequential.⁴ Fig. 1 explain the operational workflow of an automated dispensing system, from prescribing the medication to its delivery by nursing staff.^4,5 (See Table 1.)

Fig. 1.

Workflow of an Automated Dispensing System (ADS). Adapted.^4,5

Table 1.

Inclusion and exclusion criteria for the review.

Category	Inclusion Criteria	Exclusion Criteria
Study type	Peer-reviewed journal articles available in full text	Conference abstracts, editorials, and review articles
Population	Nurses working in inpatient hospital settings directly involved in medication administration; studies including other healthcare professionals only if linked to nursing workflows involving ADCs	Studies focusing solely on non-nursing staff (e.g., pharmacists or physicians) without connection to nursing medication administration
Setting	Inpatient hospital wards where individual medication doses are administered at the bedside by nursing staff	Outpatient clinics and community healthcare settings
Intervention	Implementation or use of Automated Dispensing Cabinets (ADCs) for medication distribution, including those integrated with other technologies (e.g., e-prescribing, BCMA)	Studies not involving ADCs, or focusing on other automation systems (e.g., robotic pharmacy dispensers)
Outcomes	Reported at least one nursing-related medication safety outcome (dose omissions, delayed doses, wrong-dose errors, or wrong-medication errors)	Studies not reporting these medication error outcomes
Time Frame	Published between January 2008 and 2025	Studies published before 2008
Language	English-language publications	Non-English publications

Nurses are the primary administrators of medications in inpatient hospital settings and play a central role in ensuring medication safety. Their responsibilities include verifying medication orders, selecting the correct drug and dose, administering medications at the appropriate time, and documenting administration, often within complex clinical environments characterized by high workload, time pressure, frequent interruptions, and high patient acuity.^5,6 Common medication administration errors include omissions, incorrect dose and/or drug, deviations from administration schedule and documentation errors.⁷ These errors may arise from human factors such as fatigue, distraction, and cognitive overload,⁸ as well as from system-level issues, including poorly organized medication storage, inefficient distribution processes, and limited real-time inventory control.⁹

Historically, hospitals relied on centralized medication distribution models, such as floor stock systems or unit-dose carts, in which medications were prepared in the central pharmacy and delivered to patient care units at scheduled intervals. While these models offered a degree of oversight, they were associated with delays in first-dose administration, missing doses, informal borrowing of medications between patients, and limited accountability for controlled substances.¹⁰ These inefficiencies not only posed risks to patient safety but also increased the administrative burden on nurses and reduced time available for direct patient care.¹¹

To address these challenges, healthcare systems have increasingly adopted Automated Dispensing Cabinets (ADCs), which are secure, computerized medication storage devices located in or near patient care areas.¹² First introduced in the late 1980s and widely implemented during the 1990s, ADCs were designed to decentralize medication access, improve inventory accuracy, and enhance medication safety.^13,14 Contemporary ADC systems are often integrated with electronic health records (EHRs), incorporate patient-specific drawers, support barcode verification, and generate detailed audit logs.¹⁵ These features are intended to reduce medication administration errors, improve access to urgent medications, strengthen regulatory compliance, and optimize nursing workflow.^16,17

Evidence regarding the effectiveness of ADCs in reducing medication errors is generally positive but heterogeneous. Several studies have reported reductions in specific error types, particularly omission errors and wrong-drug incidents, following ADC implementation.^18,19 Nurses frequently describe improvements in workflow efficiency, faster access to medications, and increased confidence in medication safety.²⁰ However, other studies have identified challenges associated with ADC use, including technical issues, workflow disruptions, and the emergence of unsafe workarounds that bypass safety features.^21,22 These findings suggest that the impact of ADCs on medication safety is closely linked to implementation context, staff training, system configuration, and integration with broader medication safety strategies.²³

Previous systematic reviews have examined medication management technologies in hospital settings but have focused on specific aspects of ADC use. Review⁴⁴ evaluated the effect of ADCs on omitted or delayed medication doses related to ward stock availability and reported reductions in omitted doses across nine studies. However, this review did not examine a broader range of medication administration error types or nurse-related workflow factors. Review⁴⁵ assessed multiple medication technologies, including ADCs, barcode medication administration (BCMA), and closed-loop electronic medication management systems, with a particular emphasis on controlled medication safety. While this review highlighted improvements in controlled substance management and workflow efficiency, it also reported persistent safety risks and the emergence of new error types, and it did not focus specifically on routine medication administration practices or nurse-related errors beyond controlled drugs.

These reviews provide valuable insights but do not offer a comprehensive synthesis of the overall impact of ADCs on nurse-related medication errors across inpatient hospital settings. Specifically, existing reviews have not examined a broad range of medication error types alongside contextual factors such as nursing workflow, cabinet design, system integration, and training, all of which may influence how ADCs affect medication safety in practice.

Given the widespread adoption of ADCs and the central role of nurses in inpatient medication administration, a more comprehensive evaluation of ADC-related medication errors is warranted. Although national surveys have documented high levels of ADC adoption in hospitals, evidence regarding their impact on medication safety outcomes remains inconsistent. While some studies report substantial reductions in specific error types following ADC implementation,^14,16,18,36 others have found no significant change in overall error rates or have identified new usability and workflow challenges.^9,33,35,38 Qualitative studies further highlight variability in nurses' experiences, with reports of improved efficiency and access to medications alongside accounts of temporary disruptions, queuing delays, and unsafe workarounds during early implementation phases.^29,33, 34, 35

This systematic literature review addresses these gaps by synthesizing evidence on nurse-related medication errors associated with ADC use in inpatient hospital settings. By examining a wider range of error types, study designs, data collection methods, and contextual factors, this review aims to provide a clearer understanding of how ADCs influence medication safety and nursing practice across diverse hospital contexts, and to identify key areas for future research.

2. Research questions

RQ1: What categories of nurse-related medication errors are reported in studies involving automated dispensing cabinets?

RQ2: What demographic and professional characteristics of nursing staff are reported in studies examining ADC use?

RQ3: What data collection methods are used to measure medication errors in studies involving ADCs?

RQ4: What outcomes are reported regarding the association between ADC implementation and medication errors in inpatient hospital settings?

RQ5: What knowledge gaps remain in understanding ADC-related medication errors, and how can future research better inform their safe and effective integration into clinical practice?

3. Method

The review process adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 checklist to ensure methodological rigor and transparent reporting.³⁹

3.1. Eligibility criteria

Eligibility criteria were pre-defined in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines.³⁹ Only peer-reviewed journal articles available in full text were considered for inclusion. Studies were required to investigate the use of automated dispensing cabinet (ADC) technology in hospital settings. Conference abstracts, editorials, and review articles were excluded to ensure inclusion of primary empirical evidence with sufficient methodological detail to support data extraction and analysis.

The population of interest comprised nurses working in inpatient hospital settings who were directly involved in the medication administration process. Studies examining nurses' interactions with, experiences of, or performance using ADCs were prioritized, as nurses are the primary end-users of these systems at the point of care. Studies involving other healthcare professionals, such as pharmacists or physicians, were included only when their roles were directly linked to nursing medication administration workflows involving ADCs.

Eligible studies were conducted in inpatient hospital settings, specifically on wards where individual medication doses were administered at the bedside by nursing staff. This criterion ensured that ADC use was evaluated within routine inpatient medication administration workflows. Studies conducted in outpatient clinics or community healthcare settings were excluded, as medication distribution processes in these settings differ substantially from those used in inpatient wards.

The intervention of interest was the implementation or use of automated dispensing cabinets for medication storage and distribution on hospital wards. Studies were eligible if ADCs constituted a core component of the medication dispensing process, including studies in which ADCs were integrated with other medication management technologies (e.g., electronic prescribing systems or barcode medication administration). Studies focusing exclusively on other forms of automation, such as robotic pharmacy dispensing systems, were excluded.

To operationalize nurse-related medication errors, included studies were required to report at least one medication safety outcome occurring during the medication administration stage and involving nursing staff. Specifically, eligible outcomes included dose omissions, delayed doses, wrong-dose errors, and wrong-medication errors. These outcomes were selected because they represent common administration-related error types that ADC implementation is intended to address. Studies that did not report medication administration error outcomes attributable to nursing practice were excluded.

The search was limited to studies published from January 2008 to July 2025. This time frame was selected to capture evidence reflecting the evolution and wider adoption of ADC technology, particularly following increased interoperability with electronic health record and prescribing systems in the late 2000s.^11,16 This period also aligns with the growing global emphasis on medication safety, including the World Health Organization's “Medication Without Harm” initiative launched in 2017 as part of the third Global Patient Safety Challenge.

Finally, only studies published in English were included. Non-English language publications were excluded to ensure consistency in data interpretation; this restriction is acknowledged as a potential source of language bias and is addressed in the limitations of this review.

3.2. Search strategy

To identify suitable studies, the primary author conducted a comprehensive search across four electronic repositories: PubMed, MEDLINE, Scopus, and Google Scholar, in line with PRISMA 2020 recommendations.³⁹ The search strategy was developed a priori and registered in PROSPERO (CRD420251251165), with database-specific Boolean search strings predefined in the protocol. Searches were conducted for studies published between 2008 and 2025, and were limited to English-language, peer-reviewed articles. The final search across all databases was completed in July 2025.

Given the large number of results returned by Google Scholar, the first 100 results (equivalent to 10 pages) were reviewed, consistent with prior recommendations for managing broad search engines.⁴⁰ To reduce the risk of selection bias associated with relevance-ranked search engines, a sensitivity screening was additionally performed by reviewing a further 100 results (total = 200 records), confirming that no additional eligible studies were identified beyond those captured in the initial screening. This pragmatic approach is consistent with methodological guidance and reporting practices in evidence synthesis literature, which recognize the use of fixed screening thresholds and sensitivity checks when searching Google Scholar due to its high retrieval volume and limited reproducibility.^46,48 To ensure completeness, we also examined the reference lists of included studies (backward reference list checking) and searched for studies that had cited them (forward reference list checking), an approach supported in methodological literature.⁴²

The compilation of search terms for the review was based on relevant literature and the predefined protocol. Where applicable, controlled vocabulary (e.g., Medical Subject Headings [MeSH]) was combined with free-text terms in database-specific search strategies to improve retrieval sensitivity and precision. The resultant search query was composed of three main categories of terms: (1) terms associated with Automated Dispensing Cabinets, such as “automated dispensing cabinet,” “automated medication dispensing,” and “pharmacy automation”; (2) terms linked to nurses, including “nurse,” “nursing staff,” and “healthcare professional”; and (3) terms related to medication errors, such as “medication error,” “medication administration error,” “omitted dose,” “wrong dose,” and “wrong medication.” The Boolean OR operator was used to combine terms within each category, while the Boolean AND operator was applied to combine the categories. The full, database-specific search strings for PubMed, MEDLINE, Scopus, and Google Scholar are provided in Supplementary Appendix 1 to enable full reproducibility.

Appendix 1. Database-Specific Search Strategies.

Database	Search Date	Limits/Filters Applied	Complete Search Strategy
PubMed (MEDLINE via PubMed)	July 2025	English language; publication years 2008–2025	(“Automated Dispensing Cabinets” OR “automated dispensing cabinet” OR “automated medication dispensing” OR “automated dispensing device” OR “pharmacy automation”) AND (“nurse” OR “nurses” OR “nursing” OR “nursing staff” OR “healthcare professional”) AND (“medication error” OR “medication administration error” OR “adverse drug event” OR “dispensing error” OR “prescribing error” OR “omitted dose” OR “delayed dose” OR “wrong dose” OR “wrong medication”)
MEDLINE (via EBSCOhost)	July 2025	English language; publication years 2008–2025	(“automated dispensing cabinet” OR “automated medication dispensing” OR “automated dispensing system” OR “pharmacy automation”) AND (“nurse” OR “nurses” OR “nursing staff” OR “healthcare professional”) AND (“medication error” OR “medication administration error” OR “adverse drug event” OR “dispensing error” OR “wrong dose” OR “omitted dose”)
Scopus	July 2025	English language; publication years 2008–2025;	TITLE-ABS-KEY (“Automated Dispensing Cabinets” OR “automated dispensing cabinet” OR “automated medication dispensing” OR “automated dispensing system” OR “pharmacy automation”) AND (“nurse” OR “nurses” OR “nursing” OR “nursing staff” OR “healthcare professional”) AND (“medication error” OR “medication administration error” OR “adverse drug event” OR “dispensing error” OR “prescribing error” OR “omitted dose” OR “delayed dose” OR “wrong dose” OR “wrong medication”)
Google Scholar	July 2025	English language; publication years 2008–2025; first 100 results screened	(“automated dispensing cabinet” OR “automated medication dispensing” OR “automated dispensing system” OR “automated dispensing device” OR “pharmacy automation”) AND (“nurse” OR “nurses” OR “nursing staff” OR “healthcare professional”) AND (“medication error” OR “medication administration error” OR “adverse drug event” OR “prescribing error” OR “dispensing error” OR “omitted dose” OR “delayed dose” OR “wrong dose” OR “wrong medication”)

3.3. Study selection

All retrieved records were imported into Rayyan, a web-based tool designed to support systematic review screening and collaboration.⁴³ Duplicate records were removed automatically prior to screening. The study selection process followed the PRISMA 2020 guidelines.³⁹

Two reviewers independently screened titles and abstracts to assess eligibility. Inter-reviewer agreement at the title and abstract screening stage was assessed using Cohen's kappa coefficient, demonstrating substantial agreement (κ = 0.67). Disagreements were resolved through discussion, with a third reviewer consulted when consensus could not be reached, consistent with best practice for systematic reviews.⁴¹

Full-text articles of potentially eligible studies (n = 30) were then independently assessed by the same reviewers against the predefined inclusion and exclusion criteria. Agreement at the full-text screening stage was high, with concordant decisions reached for 23 of 30 articles (76.7% agreement). Given the smaller number of full-text articles assessed, percentage agreement was used at this stage. Disagreements (n = 7) were resolved through discussion and consensus. Following full-text assessment, 14 studies met the inclusion criteria and were included. Two additional studies were identified through backward and forward reference list screening, resulting in a final total of 16 studies included in the review. The numbers of records included and excluded at each stage are presented in the PRISMA flow diagram (Fig. 2).

Fig. 2.

PRISMA flow diagram of study selection process]39[.

3.4. Data extraction

A standardized data extraction form was developed in Microsoft Excel to ensure consistency and completeness across studies. The extraction form was piloted on a subset of included studies and refined iteratively to ensure that all relevant variables were captured clearly and consistently. Two reviewers independently extracted data from all included studies. Any discrepancies identified during the extraction process were resolved through discussion and consensus, with involvement of a third reviewer when agreement could not be reached.The extracted variables are summarized in Table 2.

Table 2.

Data items extracted from included studies.

Data field	Description
Study Identification	Author(s), publication year, country of study.
Setting and Unit Type	Hospital type (e.g., tertiary, community) and specific inpatient unit (e.g., ICU, medical ward, surgical ward).
Population	Characteristics of nursing staff involved (e.g., registered nurses, licensed practical nurses) and sample size.
ADC Characteristics	Type of ADC (profiled/non-profiled), integration with EHR, barcode scanning features, patient specific drawers.
Intervention Details	Description of ADC implementation process, training provided, and duration of use before evaluation.
Outcomes	Reported medication error rates and/or types (e.g., wrong dose, omission, wrong medication), safety incidents.
Study Design	Research design (quantitative, qualitative, mixed methods) and approach to measuring errors.
Error Detection Method	How errors were identified (e.g., direct observation, incident reporting systems, chart review, surveys).
Key Findings	Summary of main results, including changes in error rates and nurse perceptions of safety/workflow.

3.5. Risk of bias assessment

The assessment was conducted independently by two reviewers, with any disagreements resolved by consensus. The Mixed Methods Appraisal Tool (MMAT)⁴⁷ was applied to evaluate the methodological quality of each included study. The MMAT is specifically designed for systematic mixed-studies reviews and critically appraises qualitative, quantitative, and mixed-methods research, yielding a quality score from 0% (lowest) to 100% (highest). Each study was first screened for clear objectives and suitable methods, and then the appropriate MMAT criteria (corresponding to its study design) were applied to assess key quality criteria according to the MMAT tool. Assessment was performed at the criterion level using the MMAT response options (“Yes”, “No”, or “Can't tell”), enabling identification of specific methodological strengths and limitations across study designs. Although MMAT developers caution against the use of aggregated numeric scores as a basis for inferential weighting or study exclusion, descriptive percentage summaries were used in this review solely to provide an overview of methodological characteristics across heterogeneous studies. Interpretation of findings was guided primarily by domain-level MMAT judgments relevant to study design, confounding (e.g, patient acuity, staffing levels, baseline error rates), and outcome measurement rather than by overall percentage scores.

4. Results

4.1. Study result

As shown in Fig. 2, a total of 184 studies were retrieved from the databases searched. After removing 25 duplicates, 159 studies remained for title and abstract screening. Of these, 129 records were excluded for reasons including irrelevance to automated dispensing cabinets, non–peer-reviewed publication type, absence of nursing involvement, outpatient setting, or lack of focus on medication errors. The full texts of 30 studies were then assessed, and 16 were excluded as they did not meet the inclusion criteria. In addition, two relevant studies were identified through backward and forward reference list checking. In total, 16 studies were included in this review.

4.2. Data synthesis

Due to the heterogeneity of study designs, settings, and outcome measures, a quantitative meta-analysis was not feasible. We therefore conducted a narrative synthesis of findings. Results were organized around the predefined research questions and key domains, including categories of nurse related medication errors, nursing staff characteristics, data collection methods, and reported outcomes of ADC implementation. Findings were compared across studies to identify common themes, differences, and trends. This approach was chosen to accommodate variability in study methodology and to provide a structured synthesis of the available evidence.

4.2.1. Characteristics of included studies

The characteristics of the included studies are presented in Table 3. The 16 studies were published between 2008 and 2025, with a noticeable increase in publications in recent years, particularly from 2020 onwards. Studies were conducted across a range of countries, reflecting diverse healthcare contexts, with the United Kingdom contributing the largest number of studies (n = 4), followed by Saudi Arabia, Taiwan, and Finland (each n = 2), while the remaining evidence was distributed across Europe, Asia, North America, and Australia. fourteen of the sixteen included studies (87.5%) were published as journal articles while two studies (12.5%) were reported as conference papers. (See Table 4, Table 5, Table 6, Table 7, Table 8.)

Table 3.

Summary of Publication characteristics (N = 16).

Characteristic	Category	Number of included publications (%)	References
Year of publication	2010	1 (6.3%)	35
	2012	1 (6.3%)	38
	2013	1 (6.3%)	34
	2014	2 (12.5%)	9,16
	2019	1 (6.3%)	28
	2020	2 (12.5%)	29,32
	2021	1 (6.3%)	33
	2023	2 (12.5%)	31,36
	2024	2 (12.5%)	27,30
	2025	3 (18.8%)	24, 25, 26
Country of study	United Kingdom	4(25%)	9., 26, 27, 31
	Saudi Arabia	2 (12.5%)	25,28
	Taiwan	2 (12.5%)	33,36
	Finland	2 (12.5%)	24,29
	Australia	1 (6.3%)	32
	Canada	1 (6.3%)	34
	France	1 (6.3%)	16,
	Spain	1 (6.3%)	38
	United States	1 (6.3%)	35
	UK & Australia	1 (6.3%)	31
Publication type	Journal article	14 (87.5%)	36, 25, 26, 27, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38
	Conference paper	2 (12.5%)	24,28
Methodological approach	Quantitative	6 (37.5%)	9,27,30,36, 37, 38
	Qualitative	3 (18.8%)	24, 25, 26
	Mixed-methods	7 (43.8%)	28,29,31, 32, 33, 34, 35
Study timeframe	Retrospective	4 (25.0%)	27,30,35,36
	Prospective	4 (25.0%)	9,16,33,38
	Not applicable	8 (50.0%)	24, 25, 26,28,29,31,32,34
Study design	Cross-sectional	5 (31.3%)	26,28,29,32,34
	Case studies	5 (31.3%)	9,24,25,31,33
	Cohort	3 (18.8%)	16,27,36
	Not reported	2 (12.4%)	30,38
	Experimental (randomized trial)	1 (6.3%)	9

Table 4.

Risk of Bias Assessment of Qualitative Studies Using MMAT.

Study No.	Author (Year)	Country	Design/Sample	Aim	Key MMAT Findings	Rating
35	Al Mutair et al. (2025)	Saudi Arabia	Qualitative (n = 15 nurses)	Explore nurses' perceptions of ADC safety	All criteria Yes	High (100%)
36	Dalby et al. (2025)	Australia	Qualitative (n = 18)	Investigate user experience of ADCs	1.5 Can't tell; others Yes	High (100%)
37	Korhonen et al. (2025)	Finland	Qualitative (n = 12)	Examine training and adaptation to ADC	All criteria Yes	High (100%)

Table 5.

Risk of Bias Assessment of Non-Randomized Quantitative Studies Using MMAT.

Study No.	Author (Year)	Country	Design/Sample	Aim	Key MMAT Findings	Rating
28	Cousein et al. (2014)	France	Quantitative, non-randomized (n = 250)	Assess the reduction in medication errors following ADC implementation	3.1–3.3 Yes; 3.4 No; 3.5 Yes	Moderate (71%)

Table 6.

Risk of Bias Assessment of Observational Quantitative Studies Using MMAT.

Study No.	Author (Year)	Country	Design/Sample	Aim	Key MMAT Findings	Rating
26	Jeffrey et al. (2024)	UK	Observational (n = 150)	Assess ADC influence on medication-error rates	3.1–3.3 Yes; 3.4 No; 3.5 Yes	Moderate (71%)
9	Cottney et al. (2014)	UK	Observational (n = 102)	Evaluate ADC intervention fidelity and outcomes	3.4 No; others Yes	Moderate (71%)
31	Rodríguez-González et al. (2012)	Spain	Observational (n = 360)	Assess error rates pre/post ADC	All criteria Yes; confounders controlled (3.4 Yes)	High (86%)
30	Said et al. (2024)	Qatar	Observational (n = 210)	Evaluate nurse satisfaction and performance post-ADC	All criteria Yes	High (100%)
26	Tu et al. (2023)	Taiwan	Observational (n = 175)	Quantify omitted-dose reduction post-ADC	3.1–3.3 Yes; 3.4 No; 3.5 Yes	Moderate (71%)

Table 7.

Risk of Bias Assessment of Non-Randomized Mixed-Methods Studies Using MMAT.

Study No.	Author (Year)	Country	Design / Sample	Aim	Key MMAT Findings	Rating
29	Metsämuuronen et al. (2020)	Finland	Mixed methods, non-randomized (n = 212)	Explore nurses' perceptions and experiences with ADC workflows	5.1–5.4 Yes; 5.5 No – limited integration reporting	Moderate (71%)
28	Elkady et al. (2019)	Egypt	Mixed methods, non-randomized (n = 180)	Evaluate ADC safety and efficiency outcomes	All criteria Yes; quantitative component met 3.1–3.5; 5.5 Yes	High (86%)
35	Wakefield et al. (2021)	USA	Mixed methods, non-randomized (n = 124)	Examine safety perceptions and workflow changes post-ADC	5.1–5.4 Yes; 5.5 Can't tell	Moderate (71%)

Table 8.

Risk of Bias Assessment of Descriptive Mixed-Methods Studies Using MMAT.

Study No.	Author (Year)	Country	Design/Sample	Aim	Key MMAT Findings	Rating
32	Craswell et al. (2024)	Australia	Mixed methods, descriptive (n = 87)	Evaluate ADC impacts on workflow efficiency and safety	5.1–5.4 Yes; 5.5 Can't tell	Moderate (71%)
31	Lichtner et al. (2023)	UK	Mixed methods, descriptive (n = 312)	Examine human factors influencing ADC utilization	All criteria Yes except 3.4 No (confounders)	High (86%)
34	Rochais et al. (2023)	Canada	Mixed methods, descriptive (n = 154)	Investigate ADC integration across hospital wards	5.1–5.4 Yes; 5.5 Can't tell	Moderate (71%)
33	Wang et al. (2021)	China	Mixed methods, descriptive (n = 96)	Explore barriers and facilitators of ADC implementation	5.1 No; 5.5 No; incomplete integration reporting	Low (43%)

With respect to methodological approach, mixed-methods designs were most frequently used, followed by quantitative and qualitative approaches. This reflects an emphasis on combining objective measures of medication errors with contextual insights into nursing practice and system use. Regarding study timeframe, a substantial proportion of studies employed cross-sectional designs, for which prospective or retrospective classification was not applicable, while the remaining studies were evenly divided between prospective and retrospective approaches. In terms of study design, observational designs predominated, including cross-sectional, cohort, and case study designs, whereas only one study adopted an experimental randomized design. Overall, these characteristics indicate considerable methodological diversity across the included studies, which should be considered when interpreting the reported effects of ADC implementation on medication error outcomes.

4.2.2. Characteristics of hospital types and care settings

As shown in chart 1, across the 16 included studies,^24–16 nearly all (15/16, 93.8%) were conducted in tertiary hospital, large teaching, or specialist hospitals/institutions that commonly integrate multiple levels of care, such as intensive care and specialty units, within the same facility.24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34^,36, 37, 38 Only one study (6.2%) was carried out in a primary hospital.³⁴ Within these tertiary settings, the implementation of automated dispensing cabinets (ADCs) covered a wide range of clinical contexts, underscoring that hospital type and clinical setting were not mutually exclusive. Chart Error! Reference source not found. demonstrates that ADCs were used in intensive care environments, including general ICUs, neonatal ICUs (NICUs), and pediatric ICUs (PICUs),^25,29, 30, 31, 32^,36 where nurses frequently managed high-risk medications. Several studies examined their application in general medical or surgical wards,^24,26, 27, 28^,33,34,37,38 while others reported deployment in specialty areas such as operating rooms and psychiatric wards.^9,29 Four studies assessed ADCs implemented across multiple care areas within the same hospital, often combining ICUs, general wards, and specialty units under one system.^{25,27,28,33,35} It should be noted that because tertiary hospitals typically function as integrated systems encompassing diverse departments ]48[, a single study often represented more than one care category. This overlap highlights that hospital type and clinical setting classifications were not mutually exclusive but instead reflected the interconnected structure of tertiary healthcare facilities where ADCs are utilized across multiple, intersecting care settings.^{9,25,29,30,32,36}

Chart 1.

(a) Hospital type. (b) Care settings

4.2.3. Risk of Bias in studies

The quality assessment of the included studies (n = 16) was conducted using the Mixed Methods Appraisal Tool,⁴⁷ which is specifically designed to evaluate studies employing qualitative, quantitative, or mixed methods designs within systematic reviews. Two reviewers independently performed the assessment, with disagreements resolved by consensus. The MMAT allows categorical responses of “Yes,” “No,” or “Can't tell,” and each domain was rated according to study design specific criteria. Scores were converted into percentages to facilitate interpretation: studies meeting ≥75% of criteria were rated High quality, those meeting 50–74% were Moderate, and those meeting <50% were Low. All individual domain ratings and corresponding overall MMAT scores were assigned based on consensus following MMAT guidance.⁴⁷ A complete list of the MMAT criteria questions (qualitative, quantitative, and mixed methods) is presented in Appendix 2.

Appendix 2. Mixed Methods Appraisal Tool (MMAT) 2018 Criteria [48].

Category of study design	Methodological quality criteria
Screening questions (for all types)	S1. Are there clear research questions?
	S2. Do the collected data allow to address the research questions?
1. Qualitative	1.1. Is the qualitative approach appropriate to answer the research question?
	1.2. Are the qualitative data collection methods adequate to address the research question?
	1.3. Are the findings adequately derived from the data?
	1.4. Is the interpretation of results sufficiently substantiated by data?
	1.5. Is there coherence between qualitative data sources, collection, analysis and interpretation?
2. Quantitative randomized controlled trials	2.1. Is randomization appropriately performed?
	2.2. Are the groups comparable at baseline?
	2.3. Are there complete outcome data?
	2.4. Are outcome assessors blinded to the intervention provided?
	2.5. Did the participants adhere to the assigned intervention?
3. Quantitative non-randomized	3.1. Are the participants representative of the target population?
	3.2. Are measurements appropriate regarding both the outcome and intervention (or exposure)?
	3.3. Are there complete outcome data?
	3.4. Are confounders accounted for in the design and analysis?
	3.5. During the study period, is the intervention administered (or exposure occurred) as intended?
4. Quantitative descriptive	4.1. Is the sampling strategy relevant to address the research question?
	4.2. Is the sample representative of the target population?
	4.3. Are the measurements appropriate?
	4.4. Is the risk of nonresponse bias low?
	4.5. Is the statistical analysis appropriate to answer the research question?
5. Mixed methods	5.1. Is there an adequate rationale for using a mixed methods design to address the research question?
	5.2. Are the different components of the study effectively integrated to answer the research question?
	5.3. Are the outputs of the integration of qualitative and quantitative components adequately interpreted?
	5.4. Are divergences and inconsistencies between quantitative and qualitative results adequately addressed?
	5.5. Do the different components of the study adhere to the quality criteria of each tradition of the methods involved?

4.2.4. Qualitative studies

All three qualitative studies (34, 35, 36) received high quality ratings. Each study articulated clear research questions (criterion 1.1: “Are the research questions clear and adequately addressed?” = yes), used suitable qualitative data-collection approaches (criterion 1.2: “Are the data collection methods appropriate to answer the research question?” = yes), and demonstrated strong alignment between data, interpretation, and conclusions (criteria 1.3–1.5: “Are the findings adequately derived from the data?”, “is the interpretation of results sufficiently supported by the data?”, and “is there coherence between data sources, collection, analysis and interpretation?” = yes). Only one study reported minor uncertainty in coherence (criterion 1.5 = can't tell).

4.2.5. Quantitative studies

All six quantitative studies^{9,26,29,31,33,37} demonstrated moderate to high methodological quality according to the⁴⁷ criteria. Each clearly defined its objectives and used appropriate measurement methods (criterion 3.1: “Are participants representative of the target population?”, criterion 3.2: “Are measurements appropriate regarding exposure/intervention and outcome?”, and criterion 3.3: “Are there complete outcome data?” = yes). Most studies also ensured that the intervention was administered as intended (criterion 3.5: “Was the intervention administered (or exposure occurred) as intended?” = yes). However, the main limitation across several studies was insufficient control for confounding (criterion 3.4: “Are confounders accounted for in the design and analysis?” = no), which affected the overall quality of some papers.

• (a)Non-randomized quantitative study (n = 1)

The single non-randomized study²⁸ evaluated the effect of ADC implementation on medication-error rates using a pre–post comparison design. The study met most MMAT criteria, including representativeness of participants (criterion 3.1 = yes), appropriateness of measurements (criterion 3.2 = yes), and completeness of outcome data (criterion 3.3 = yes). However, it did not control for confounding variables (criterion 3.4 = no), thereby slightly reducing internal validity. Overall, the study achieved a Moderate quality rating.

(b) Observational quantitative studies (n = 5)

The remaining five quantitative studies^{16,26,29,31,33} were observational in design and focused on the impact of ADC systems on workflow efficiency, nurse satisfaction, and medication safety. All presented representative samples and reliable outcome measures (criteria 3.1–3.3 = yes). Three of the five studies controlled for consistent intervention fidelity (criterion 3.5 = yes), though several did not control for confounders (criterion 3.4 = no), which led to moderate quality rankings. While studies,^16,29however, provided explicit controls for confounders, leading us to rate them high in terms of methodological rigor.

4.2.6. Mixed-methods studies

The seven mixed-methods studies ]24, 25, 27, 29, 31, 33, 39[were determined to be of a generally high methodological quality. Five studies achieved high ratings (≥ 75%) due to clearly justified mixed designs (criterion 5.1: “Is there an adequate rationale for using a mixed-methods design?” = yes), effective integration of qualitative and quantitative findings (criteria 5.2–5.3: “Are the different study components effectively integrated to address the research question?” and “Are the outputs of integration adequately interpreted?” = yes), and explicit treatment of divergences (criterion 5.4: “Are divergences and inconsistencies between quantitative and qualitative results adequately addressed?” = yes). Two studies were rated moderate because of limited integration detail or missing justification for using a mixed approach (criterion 5.1: “Is there an adequate rationale for using a mixed methods design to address the research question?” = can't tell) and (criterion 5.5: “Do the different components adhere to the quality criteria of each tradition involved?” = can't tell), while study]38[received a low score due to incomplete reporting and lack of a clear rationale for employing a mixed-methods design.

• (a)Non-randomized mixed-methods studies (n = 3)

Studies^24,25,32 assessed the impact of ADC introduction on safety and workflow outcomes using both qualitative interviews and quantitative data from non-randomized designs. All three justified their mixed design (criterion 5.1 = yes) and demonstrated good integration of findings (criteria 5.2–5.3 = yes); however, studies ]24[ and ]32[ did not fully address divergences between components or discuss data-integration limitations (criteria 5.4–5.5 = can't tell), resulting in moderate ratings. Study ]25[ achieved high quality due to its comprehensive integration and full reporting of both methodological components.

• (b)Descriptive mixed-methods studies (n = 4)

The remaining four mixed-methods studies [27,9,29,37] used descriptive designs to explore nurse experiences and organizational outcomes after ADC implementation. All studies were found to comply with most MMAT domains, particularly criterion 5.2 (integration of components) and criterion 5.3 (interpretation of combined outputs), indicating successful merging of qualitative and quantitative data. Studies]27[and]30[however, lacked a detailed explanation of the rationale for mixing (criterion 5.1 = can't tell) and did not confirm adherence to separate methodological quality standards (criterion 5.5 = can't tell). Study]38[was rated low as several details were missing, while study]28[achieved a high quality rating.

Across all 16 studies, 13 (81.3%) achieved ≥70% of MMAT criteria (high quality), 2 (12.5%) were rated moderate (50–69%), and 1 (6.2%) was rated low (<50%). The most frequent methodological weaknesses were insufficient control for confounding factors in quantitative designs and limited integration reporting in mixed-methods studies. While descriptive summaries indicated that several studies met multiple MMAT criteria, domain-level appraisal identified recurring methodological limitations, including the predominance of observational designs, limited adjustment for confounding variables, and reliance on incident reporting systems or self-reported data, which may be subject to underreporting and measurement bias. Therefore, despite the generally acceptable methodological quality of these studies, a degree of caution must be applied when interpreting the evidence due to design and measurement-related limitations.

Accordingly, observed reductions in reported medication error rates should be interpreted as associations rather than definitive evidence of the effectiveness of automated dispensing cabinets, as the predominance of observational designs and limited confounding control substantially restrict causal inference. All assessments were conducted by dual independent reviewers in accordance with the 2018 MMAT guideline.⁴⁷

4.2.7. Characteristics of population

The characteristics of nurse participants varied notably across the included studies, as summarized in Table 9. Seven studies (43.8%) focused exclusively on staff or bedside nurses,^{24,28,29,31,32,34,38} reflecting the central role of front-line nursing staff in medication administration and ADC operation. Three studies (18.8%) recruited nurse managers or directors,^25,31,35 emphasizing leadership perspectives on workflow oversight, controlled drug governance, and system accountability. Meanwhile, two studies (12.5%) featured mixed nursing cohorts comprising both bedside and managerial nurses,^24,34 offering a broader view that bridged operational and administrative experiences. Overall, bedside nurses represented the dominant study population, consistent with their direct interaction with ADCs during daily medication rounds, whereas managerial participation remained comparatively limited.

Table 9.

Population and sample characteristics.

Characteristic	Details	Number of studies (%)	References
Profession	Staff/bedside nurses	7 (43.8%)	24,29,30,32,33,35,39
	Nurse managers/directors	3 (18.8%)	25,32,36
	Mixed nurses (staff + management)	2 (12.5%)	24,35
Nursing experience	Majority senior (e.g., Staff Nurse I, >15 yrs)	2 (12.5%)	29,30
	Nurse managers (leadership roles)	2 (12.5%)	25,32
	Not reported	12 (75%)	24,26, 27, 28,31,33, 34, 35, 36, 37, 38, 39
Sample size	Largest (312 nurses)	1 (6.3%)	29
	100–200 nurses	2 (12.5%)	29,35
	50–99 nurses	1 (6.3%)	30
	<20 nurses	3 (18.8%)	24,25,32
	Not reported	9 (56.3%)	26, 27, 28,31,34,36, 37, 38, 39

Details regarding nursing experience were inconsistently reported. Only two studies (12.5%)^24,34 specified participants' seniority, while twelve (75%) did not provide explicit data on years of practice or professional grade.^24,26,27,30,32, 33, 34, 35, 36, 37, 38 Two studies (12.5%) involved predominantly senior bedside nurses, often at the Staff Nurse I level with more than fifteen years of clinical experience,^28,29 whereas two others (12.5%) targeted nurse managers in leadership roles.^25,31 No studies focused primarily on junior or newly qualified nurses, suggesting that ADC research to date has largely represented experienced professionals who are familiar with hospital medication workflows and safety protocols.

Sample sizes varied substantially across studies, reflecting the methodological diversity of the evidence base. The largest investigation surveyed 312 bedside nurses,²⁸ while two studies (12.5%) recruited 100–200 nurses.^28,34 One medium-scale study included 50–99 nurses,³⁹ while three smaller qualitative studies involved fewer than twenty participants,^24,25,31 often through focused interviews or observation sessions. More than half of the studies (56.3%) did not report the exact number of nurse participants,26, 27, 28^,30,33,35, 36, 37, 38 as they relied on administrative datasets, ADC transaction logs, or incident reports rather than direct staff participation. Collectively, these findings demonstrate that most ADC research engaged bedside nurses in small to moderate sample sizes, with limited reporting on participant experience levels, underscoring the need for greater methodological transparency in future investigations.

Reporting of participant characteristics varied considerably across studies. While most studies identified participants as registered nurses, detailed information on professional seniority, years of experience, or employment status was frequently absent. In particular, junior nurses as well as temporary or agency staff, were rarely reported as distinct groups, limiting the ability to assess how ADC-related medication errors may differ across nursing experience levels. As a result, findings should not be interpreted as representative of all nursing populations.

4.2.8. Characteristics of automated dispensing cabinets (ADCs)

The scope of ADC deployment varied considerably across the included studies, as summarized in Table 10. Six studies (37.5%) were conducted in hospitals with small-scale implementations (≤10 units),^28,32,35, 36, 37, 38 including sites with a single cabinet on one ward or fewer than 10 units within a specialty department. Another six studies (37.5%) reported medium-scale rollouts of approximately 11–30 cabinets.24, 25, 26^,30,32,33 Large-scale deployments (≥30 units) were described in four studies (25.0%),^24,28,29,38 with some hospitals installing over 40 units across multiple departments. The maturity of ADC implementations also differed. Two studies (12.5%) examined newly introduced systems (<1 year in operation),^28,35 while most evaluated ADCs in place for 1–3 years (7/16, 43.8%).25, 26, 27, 28, 29^,33,36 Five studies (31.3%) investigated mature systems 3–5 years post-installation,^{24,31,32,35,38} and two (12.5%) assessed ADCs running for over 5 years.^32,38

Table 10.

ADC implementation scale, history, and technical features.

Feature	Description/Category	Number of studies (%)	References
Number of ADC units	Small-scale (≤10 units)	6 (37.5%)	27,31,34, 35, 36, 37
	Medium-scale (11–30 units)	6 (37.5%)	24, 25, 26,31,33,34
	Large-scale (≥31 units)	4 (25%)	24,29,30,39
	Mean (standard deviation)	5.33	24, 25, 26,28,31, 32, 33, 34, 35, 36, 37, 38
	Range	4–6

Feature	Description/Category	Number of studies (%)	References
ADC implementation history	<1 year (very new)	2 (12.5%)	27,35
	1–3 years	7 (43.8%)	25, 26, 27,29,30,34,37
	3–5 years	5 (31.3%)	24,31,32,35,38
	>5 years	2 (12.5%)	33,38
	Mean (standard deviation)	4	16,24,30,31,34
	Range	2–7
Key technical features	Biometric or unique login for each user	3 (18.8%)	28,31,36
	Controlled drug safeguards (dual login, counts)	2 (12.5%)	,35,38
	Barcode scanning at administration (BCMA)	1 (6.3%)	38
	Electronic audit trail & alerts	2 (12.5%)	30,31
	Restricted removal timing windows	1 (6.3%)	36
	Partial formulary in ADC (not all meds)	2 (12.5%)	16,36

Common technical features included secure individual user logins, often with biometric authentication for user access tracking and automatic electronic documentation of all transactions.^28,29,36 Many ADCs incorporated safety alerts and access restrictions; for example, one ICU system restricted medication removal to within six hours of scheduled administration to prevent early or late dosing.³⁶ Controlled substance management was a consistent focus, with most sites stocking narcotics under strict protocols such as blind counts, dual nurse authentication, and automated discrepancy alerts.^31,35 Some hospitals excluded certain medications from ADC storage due to cabinet capacity, regulatory limits, or low use frequency.^16,36 Hardware configurations typically included multiple drawers or compartments, with some installations featuring refrigerated modules or auxiliary cabinets for large-volume items. User interfaces supported on-screen medication selection and, in profiled systems, patient-specific lists. Not all ADCs integrated with bedside barcode medication administration (BCMA), and lack of BCMA capability was noted as a safety limitation, particularly due to reduced verification at the point of administration. Override functions were available in several systems; one study reported 14% of controlled drug transactions as overrides, raising concern regarding potential misuse. Overall, ADC implementations demonstrated controlled and auditable medication dispensing processes, although the sophistication of features and integration with hospital systems varied notably between sites.

While ADC characteristics were primarily reported descriptively, several included studies suggested links between specific system features and medication safety outcomes. Studies describing profiled ADCs and integration with BCMA reported reductions in selection-related errors, including wrong drug and wrong dose incidents, particularly in inpatient and critical care settings.^{11,14,18,36,38} These findings were commonly attributed to improved verification at the point of medication access and administration. In contrast, studies examining override functionality highlighted potential safety risks when overrides were frequently used outside urgent clinical situations, including increased opportunities for selection errors and reduced effectiveness of built-in safety checks.^21,22,31 Qualitative and observational studies further emphasized that inappropriate override use was often linked to workflow pressures, system configuration issues, and gaps in user training.^29,31,34 Although these associations do not establish causality, they indicate that the impact of ADCs on medication safety depends not only on system availability but also on how specific features are configured, governed, and used in clinical practice.

4.2.9. Data collection methods

As shown in Chart 2, the included studies employed diverse data collection approaches, often aligned with their study design. Direct observation of medication-related processes was the most commonly used method (6/16, 37.5%), with trained observers monitoring nurses during medication administration to document workflow and errors.^{24,27,29,31,37,38} In some cases, observations were conducted covertly to minimize behavioral changes during data collection.³⁸ While observational methods provide detailed insight into real-world practice, they may still be subject to observer or Hawthorne effects, potentially influencing staff behavior and leading to underestimation of routine workarounds or errors.

Chart 2.

Number of studies using each data-collection method in the included studies (multiple methods per study possible).

Surveys or questionnaires were used in 4/16 studies (25%) to assess nurse perceptions, satisfaction, and perceived impacts on workflow and safety.^28,29,31,34 Large-scale examples included an online survey involving 312 nurses²⁸ and a departmental survey of 81 nurses.²⁹ Although survey-based methods offer valuable insight into user experience and perceived safety benefits, they rely on self-reported data and do not directly measure medication error rates.

Interviews and focus group methods were reported in 4/16 studies (25%), capturing in-depth perspectives from nurses, pharmacists, and managers on ADC implementation challenges and facilitators.24, 25, 26^,35 These qualitative approaches provided contextual understanding of system use and workflow integration but were not designed to quantify error frequency or direction of effect. Retrospective record reviews were employed in 3/16 studies (18.8%), analyzing medication incident reports, ADC transaction logs, or electronic health record data to assess error rates and usage trends.^27,30,36 However, reliance on incident reporting systems may underestimate true error rates due to known underreporting and variability in reporting practices. A small number of studies (3/16, 18.8%) employed mixed-methods designs, combining qualitative insights with quantitative measures to provide a more comprehensive view of ADC use and its perceived and measured impacts.^24,29,33

4.2.10. Characteristics of medications

Among the 16 included studies, 6 (37.5%)^{16,27,28,30,36,38} quantitatively measured medication error rates or related outcomes. As shown in Chart 3 the most frequently examined error types were dose omissions and dosing errors. Two studies^16,27 specifically focused on omitted dose errors: one reported a marked reduction in omissions due to improved drug availability after ADC installation,²⁷ while another found omissions remained the most common error type pre and post implementation, despite an overall error frequency decline.¹⁶ Other studies assessed a wider range of administration errors, including wrong dose, wrong drug, and wrong time errors. For example, in a geriatric unit, significant reductions in wrong dose and wrong drug errors were observed after transitioning to a unit dose ADC system.¹⁶ An ICU based study reported improved dispensing accuracy, with no serious harm errors recorded under the ADC system.³⁶ A randomized controlled trial in a psychiatric ward found lower overall medication error rates with ADC use compared to traditional carts.⁹

Chart 3.

Frequency of reported medication-error outcomes and controlled-drug considerations across included studies.

The majority of studies (10/16, 62.5%) did not provide quantitative error rates, instead reporting qualitative or process focused findings. In survey-based studies, most nurses agreed that medication administration felt safer and more reliable with ADCs,^28,34 even without objective error data. Controlled substance management was a recurrent theme. Several studies described enhanced narcotic governance through ADC features such as biometric access, dual authentication, and continuous audit logs.^31,34 In most implementations, controlled drugs were stocked in ADCs under strict safeguards. Nurses frequently reported feeling more secure when handling narcotics due to improved oversight and reduced risk of diversion.^28,34 Some sites observed the near elimination of missing opioid doses once dispensed exclusively via ADCs with tracking.²⁸ Only one study excluded controlled substances from ADC storage due to regulatory and design limitations.³⁶ None of the included studies compared error rates between controlled and non-controlled drugs, with most discussions focusing on security and accountability improvements rather than differential error frequencies.

4.2.11. Nurses' perception of ADCs

As shown in Chart 4, across most of the included studies (12/16, 75%), nurses expressed positive perceptions of ADC implementation, with many emphasizing improvements in workflow efficiency and patient safety. They frequently described how ADCs enabled faster access to medications, reduced unnecessary trips to central pharmacies, and minimized delays in patient care, which in turn allowed them to spend more time at the bedside.^{16,24,25,29,34,35} In one large survey, as many as 90% of nurses agreed that ADCs made their work easier overall, underscoring the extent to which these systems were viewed as supportive tools rather than burdens.³⁴

Chart 4.

Nurses' perception of adcs.

Although the overall tone was favourable, a few studies (3/16, 18.8%) noted minor challenges during the early phases of implementation. Nurses sometimes reported queuing at cabinets during peak times,³² while others described initial increases in preparation time as they adjusted to new processes.³³ However, these difficulties were consistently framed as temporary, with most reports indicating that they resolved within a few months of continued use, suggesting that familiarity and adaptation were key to overcoming early frustrations. Concerns about unsafe practices were limited, as studies (6/16, 37.5%) explicitly confirmed the absence of major workarounds.^24,26,28,34, 35, 36 Where deviations did occur, they were generally minor, such as the use of shared logins or pre-staging of medication removals.^28,32 Importantly, nurses often contrasted these occasional lapses with the broader sense of reassurance that ADCs provided in managing controlled substances.

Features such as biometric access, automated logging, and dual authentication were consistently highlighted as safeguards that enhanced accountability.^28,31,34 Training was another central theme shaping nurses' experiences. More than half of the studies (9/16, 56.3%) described structured training programs, which often blended vendor led sessions with support from super users who provided peer to peer mentorship.24, 25, 26^,29,32 This structured approach was strongly associated with confidence in daily practice, and in one survey involving over 300 nurses, only a single respondent reported that the training was insufficient.²⁸ Several studies went further to emphasize the importance of refresher courses and ongoing support, noting that continued education was essential to maintain competency and adapt to system upgrades.^25,27 Nurses often described training not as a one-time event but as an ongoing process that reinforced their trust in the system.

Taken together, these findings depict nurses as generally supportive of ADCs, viewing them as valuable tools that enhanced efficiency, safety, and accountability. While initial challenges and minor deviations were acknowledged, they were largely overcome through experience, robust training, and the system's built-in security features, leaving nurses more confident in their ability to provide safe and timely care.

5. Discussion

5.1. Principal findings

This systematic review synthesizes evidence on the relationship between automated dispensing cabinet (ADC) implementation and nurse-related medication errors in inpatient hospital settings. Overall, the included studies indicate that ADC use is associated with reductions in certain reported medication error types, particularly omitted doses and selection-related errors such as wrong drug or wrong dose incidents.^8,16,27,37 These associations were most frequently observed in settings where ADCs were integrated into routine nursing workflows and supported by complementary safety mechanisms.

At the same time, the evidence base does not demonstrate uniform effects across all medication error categories. Improvements were more consistently reported for omission-related outcomes than for timing-related errors, and notably, none of the included studies provided quantitative measures of delayed dose administration.^24,38 This suggests that while ADCs are widely perceived to improve medication availability and workflow efficiency, empirical evidence on their impact on administration timeliness remains limited. Much of the existing literature prioritizes process indicators, descriptive audits, or staff perceptions rather than standardized error rate measurement.

Qualitative and mixed-methods studies consistently reported positive nurse perceptions of ADCs, including improved workflow organization, faster access to medications, and greater confidence in medication safety.^28,29,34 Nurses frequently described ADCs as enhancing accountability and control, particularly for high-alert and controlled medications, through secure access and transaction tracking.^30,31 However, these findings must be interpreted with caution. Self-reported perceptions may be influenced by social desirability bias or positive attitudes toward newly introduced technologies, especially in early implementation phases. As such, perceived safety improvements should not be equated with objectively verified reductions in medication error occurrence.

Importantly, although many included studies met multiple MMAT criteria and were categorized as high quality, this should not be interpreted as evidence of low risk of bias or strong causal inference. The predominance of non-randomized designs, limited adjustment for confounding variables, and frequent reliance on incident reporting systems substantially constrain causal interpretation. As noted in medication safety research more broadly, changes in reported error rates following technology implementation may reflect shifts in reporting behavior, documentation practices, or workflow adaptation rather than true changes in underlying error frequency.^2,4 Consequently, ADCs should be understood as supportive technologies associated with improved safety-related processes, rather than as standalone interventions that directly reduce medication errors.

5.2. Practical and research implications

The findings of this review suggest several practical considerations for hospitals and nursing leadership when implementing or optimizing automated dispensing cabinets in inpatient settings. While ADC use was commonly associated with improvements in reported medication safety and workflow efficiency, these benefits were not uniform across studies and appeared to depend strongly on how the technology was implemented and integrated into daily practice. This underscores the importance of viewing ADCs as one component of a broader medication safety strategy rather than as a standalone solution.

From a practical perspective, successful ADC implementation requires careful alignment with existing medication workflows. Studies consistently indicated that ADCs function more effectively when integrated with complementary systems, such as electronic medication records and bedside barcode medication administration, and when cabinet configuration reflects actual ward-level medication needs. Adequate planning around cabinet placement, access points, and inventory selection may help reduce bottlenecks and discourage unsafe workarounds, particularly during peak medication administration times. Ongoing training and local support are also critical, as early implementation challenges were often linked to unfamiliarity with system functions or incomplete understanding of safety features such as profiling and override controls.

Importantly, nurses' experiences highlighted that technology alone does not eliminate the need for professional judgment and vigilance. While ADCs can support safer medication practices by improving access control and traceability, their effectiveness depends on consistent and appropriate use. Engagement of frontline nursing staff in system configuration, review of override practices, and continuous quality improvement activities may help sustain safe use over time and mitigate the emergence of unintended risks.

From a research perspective, the findings point to several areas requiring further investigation. Most available evidence is derived from observational studies conducted in single hospitals or units, limiting the ability to draw firm conclusions about effectiveness across diverse settings. Future studies should employ stronger study designs, including controlled and longitudinal evaluations, to better assess how ADCs influence medication error rates over time. There is also a need for greater standardization in outcome definitions and measurement methods, particularly for medication administration timing and delayed doses, which were rarely quantified despite being frequently discussed by nursing staff.

Future research should give greater attention to underrepresented nursing populations, such as junior nurses, temporary staff, and newly trained personnel, whose experiences with ADCs remain largely unexplored. Finally, examining how specific system features and local implementation decisions interact with organizational culture and workload pressures may provide more actionable insights for hospitals seeking to maximize the safety benefits of ADC technology.

5.3. Limitations

This review has several important limitations that should be considered when interpreting the findings. First, participant characteristics were incompletely reported across many included studies. Information on nurse seniority, years of experience, or employment status was often absent, and data on junior nurses, newly qualified staff, or temporary personnel were rarely provided. This limits assessment of representativeness and highlights a notable evidence gap, as interactions with medication technologies and vulnerability to medication errors may differ according to experience level and staffing role.

Second, there was substantial heterogeneity across studies in terms of design, outcome definitions, and methods used to identify medication errors. Medication error outcomes were variably defined and measured, with limited standardization across studies. This heterogeneity constrained meaningful comparison and synthesis and reduced confidence in drawing consistent conclusions across settings.

Third, many studies relied on incident reporting systems or self-reported survey data to identify medication errors. These approaches are known to be susceptible to underreporting, reporting bias, and changes in reporting behavior following system implementation. Direct observation studies, while providing valuable insights into workflow and error processes, may also be influenced by observer effects. Collectively, these limitations mean that observed changes in reported error rates may not fully reflect true changes in underlying error occurrence.

Fourth, the majority of included studies were conducted in tertiary hospitals and high-income healthcare settings, which limits the generalizability of findings to smaller hospitals, resource-limited settings, or healthcare systems with different staffing models and technological infrastructure. Caution is therefore required when extrapolating these findings to broader contexts.

Finally, although comprehensive database searching and reference list screening were undertaken, the possibility of publication bias cannot be excluded. Studies reporting neutral or negative findings may be underrepresented in the published literature, potentially inflating perceptions of benefit. In addition, the review was limited to English-language publications, which may further contribute to selection bias.

These limitations underscore the need for cautious interpretation of the evidence. While the findings suggest that ADC implementation is associated with improvements in certain medication safety processes, the strength of inference is constrained by methodological variability, incomplete reporting, and limited generalizability.

6. Conclusion

Automated dispensing cabinets represent an important technological development in hospital medication management and are widely implemented in inpatient settings. The evidence synthesized in this review suggests that ADC use is commonly associated with improvements in workflow organization, controlled substance management, and nurses' perceived confidence in medication safety. However, these findings should be interpreted cautiously, as the available evidence is heterogeneous and largely derived from observational studies with variable outcome definitions and measurement approaches.

This review also highlights that ADCs are not standalone solutions for medication safety. Their effectiveness appears to depend heavily on local implementation context, including system configuration, staff training, integration with other health information technologies, and broader organizational culture. Inadequate planning, limited user training, or misalignment with existing workflows may reduce potential benefits and contribute to workarounds or unintended risks.

Given these considerations, ADCs should be viewed as supportive components of broader medication safety strategies rather than as interventions that independently reduce medication errors. Future research should focus on using standardized definitions of medication errors, applying stronger and more controlled study designs, and incorporating longer follow-up periods to better assess sustained effects. In addition, greater attention is needed to underrepresented nursing populations, such as junior nurses, newly qualified staff, and temporary personnel, whose interactions with ADC systems remain insufficiently studied. Addressing these gaps will help clarify the conditions under which ADCs can most effectively support safe medication administration in diverse hospital settings.

CRediT authorship contribution statement

Arab Elsadig Salih Osman: Writing – review & editing, Writing – original draft, Methodology. Nour Isleem: Formal analysis. Dena Al-Thani: Supervision. Aya Elsakka: Methodology.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.