Registration

The protocol was registered and is available on the International Prospective Register of Systematic Reviews (PROSPERO) at the Centre for Reviews and Dissemination, University of York, United Kingdom. [PROSPERO 2023 CRD42023390727).

Study design

A systematic review of the international literature was conducted and reported as per the PRISMA (The Preferred Reporting Items for Systematic reviews and Meta-Analyses) 2020 Statement [25].

Eligibility criteria

The inclusion criteria were kept deliberately broad to identify all relevant studies. The inclusion criteria were conceptualised using the PECOS (Population, Exposure, Comparison, Outcome, Study Design) framework, as follows:

1. –. P: Adult and paediatric patients in a community pharmacy setting.
2. –. E: Over the counter or prescription medicines
3. –. C: Not applicable

O: Medication errors/safety problems, including those related to: (1) prescribing, (2) transcribing and documenting, (3) dispensing, (4) administering, and (5) monitoring.

S: Quantitative studies

Reviews, letters, editorials, commentaries, qualitative and intervention studies (i.e., randomised controlled trials, non-randomised controlled trials, controlled before and after studies looking to report the effectiveness of an intervention to promote safety) were excluded. Articles were not excluded from the systematic review based on study quality. Language restrictions were not applied. Studies were excluded if published before January 1995. The choice of the ‘January 1995’ date was made to include a sufficient breadth of research conducted both preceding and after the publication of the 1999 IOM report.

A community pharmacy setting was defined as ‘a healthcare facility that provides pharmaceutical and cognitive services to the community’ [26].

Errors were categorised as per Table 1 below. The severity of medication error referred to the potential harm or actual harm associated with medication errors [30].

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Error type and categorisation.

https://doi.org/10.1371/journal.pone.0322392.t001

Data sources and search strategy

A systematic search was conducted through the following databases and search engines from 1 January 1995 until 31 December 2023. The following databases were searched: MEDLINE (Ovid), Embase (Ovid), Cochrane Central Register of Controlled Trials, ISI Web of Science, Scopus, Database of Abstracts of Reviews of Effects (DARE), Health System Evidence, Global Health Database, Joanna Briggs Institute Evidence-Based Practice Database, Academic Search Complete, ProQuest Dissertations, PROSPERO, Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCO), ScienceDirect (Elsevier), Health Management Information Consortium (HMIC), Eastern Mediterranean Regional Office of the World Health Organization (WHO) (EMRO), and Google Scholar. Manual searching of the bibliographies of key articles and other review articles was also undertaken. Unpublished studies (i.e., grey literature) were identified through abstracts of conference proceedings and dissertation abstracts. The search also included theses.com and ProQuest as data sources as they provide access to doctoral dissertations, master’s theses, and other research works that may not be published in peer-reviewed journals.

Filters and advanced search strategies were applied according to specific databases. There were no restrictions imposed on the age or specific groups of patient populations that were included.

To locate relevant studies, search terms were chosen from different categories related to the systematic review PECOS. (Supplementary file 1 outlines the search strategy for each database).

Study selection

Endnote^© was used to remove duplicate titles from the search. Titles and abstracts were then screened for eligibility by a single researcher (MH: Maguy El Hajj) using the inclusion criteria. Eligible studies were exported to Rayyan^© software where full-text screening was undertaken by a single researcher (MH). Rayyan^© is a free, user-friendly, web-based software designed to facilitate the screening and selection of studies. It enables comparison of inclusion and exclusion decisions, making the review process more efficient [31]. Studies were included if they satisfied the pre-specified inclusion criteria. Authors were contacted by email to request missing information. If no response was received after one week, the study was excluded from inclusion.

In case of doubts regarding the inclusion of articles, discussions were undertaken with the senior authors (AH: Andy Husband and AT: Adam Todd) who had consensus. The abstracts of articles published in languages other than English were translated into English using Google Translate® and screened for eligibility. Native speakers were consulted to translate studies in German, Dutch and Portuguese if considered relevant for a full-text review.

Data extraction

A structured tool for extracting data was developed, incorporating the following areas: bibliographic details (author, year of publication, DOI), study design, study setting, ‘type’ of error in the medication use process (prescribing, transcribing and documenting, dispensing, administering, monitoring), year and duration of data collection, approach used for identifying error, main findings, limitations, and conclusions. The tool was piloted tested and refined on five randomly chosen included studies. The data were extracted by one researcher (MH) and checked in full by a second researcher (RA: Rayah Assiri) between March and May 2024. Any disagreements were discussed with the senior authors (AH and AT) who had consensus. Missing data was addressed by contacting the corresponding author of the study. If no response was received, the information was recorded as unavailable.

Quality assessment

The quality of each included study was assessed by a single researcher (MH) using a risk of bias tool for medication errors previously designed and used by Campbell et al., in a systematic review on medication errors in community pharmacies in the United States [22]. This tool assesses four sources of bias: 1) selection bias 2) identification bias 3) error categorisation bias and 4) conflict of interest bias [22].

Data synthesis and analysis

Textual summaries and tables were created to align with the primary studies’ outcomes and characteristics, utilising the extracted data including country of the study, study design, number of included pharmacies, year of data collection, error detection method and responsible person, safety hazard, population studied, inclusion and exclusion criteria, rate of errors, type, nature and seriousness of errors, implicated medication(s) and limitations. The study results were systematically organised by geographical region in accordance with the 2023 World Bank classification system [32]. Narrative and descriptive synthesis of study results was used. The strategy for narrative synthesis was conducted focusing on four elements: ‘theory development, developing a preliminary synthesis, exploring relationships within and between studies, and assessing the robustness of the synthesis’ [33].

Studies characteristics

The characteristics of included studies are presented in Table 2. These studies were systematically organised by geographical region in accordance with the World Bank Classification System. [32].The predominant location of studies was in Europe and Central Asia (n = 27 studies) [24, 28,38–62], followed by North America (n = 20) [76–95]. Ten studies were conducted in the Middle East and North Africa (MENA) [66–75], while six studies were from South Asia [96–101]. Four studies were conducted in East Asia and Pacific [34–37]. Three studies were from Latin America and the Caribbean regions [63–65] and from the Sub-Saharan Africa region [102–104].

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics of Included Studies*.

https://doi.org/10.1371/journal.pone.0322392.t002

The majority of the included studies (n = 67) were published in English [24,28,34–44,46,47,49,50,53–62,64,66–104]. Articles published in other languages included two studies published in Spanish [51,52], two in Portuguese [63,65], one in Dutch [45] and another in Danish [48]. These publications underwent translation into the English language.

The number of included community pharmacies varied across the studies and countries, ranging from one pharmacy in nine studies [45,49,78,85,93,96,63,65,104] to 2,856 pharmacies in a single study [77].

The studies varied in the duration of data collection, ranging from three days [41,42], to 10 years [80].

The predominant study design was cross-sectional (n = 20) [36,40,52,55,63,65,67,68,70,73,76,82,83,87,96,97,100,102–104]. Thirteen studies employed a retrospective design [24,47,53,77,81,89–92,94,98,37,64], while another thirteen studies were prospective [42,43,51,58,85,86,88,69,72,74,75,34,35]. Conversely, information on the study design was not provided in seventeen studies [41,44,46,48,49,54,56,57,28,59–62,79,80,95,101].

Various approaches were used to identify medication errors. Twenty-eight studies opted for a review of prescriptions [28,37,42,44,48,53,60,63–67,70,73,84,85,88,89,93,95–102,104]. Twenty-eight studies relied on self-reported errors or incidents [24,34,36,38,39,41,45–47,49–51,55–59,61,62,74,75,77–79,81,90–92]. Additionally,thirteen studies utilised direct observation techniques to detect medication errors [39,52,54,68,69,71,72,76,82,83,86,87,94]. In one study, patients’ best possible medication history (BPMH) were reviewed versus their community pharmacy-prepared dose administration aids [35]. While another study compiled Quality related events (QREs) occurring in community pharmacies and published in the Institute for Safe Medication Practices (ISMP) Canada safety bulletin newsletters [80].

Fifty-three studies did not specify the characteristics of the enrolled patients [24,38–41,44–51,28,55–62,78,81–87,89–95,80,69–75,96–99,101,36,64,65]. Six studies included patients without imposing any restrictions on age or sex [34,37,43,63,77,100]. Two studies exclusively focused on paediatric patients [53,67]. Additionally, three studies enrolled patients with respiratory diseases or those using inhalers[52,54,76]. Furthermore, one study specifically targeted patients with Parkinson’s disease who were treated with levodopa/carbidopa [88], while another study focused on patients receiving anti-retroviral agents [104]. One study only included geriatric patients using community pharmacy-prepared dose administration aids [35]. And another study targeted patients from the private health sector except pregnant and psychiatry patients [102].

Fifty- three studies focused exclusively on prescribed medications [28,35–39,41,44,47–49,53,54,56–58,60–67,70–76,82–98,100–104] while 16 studies did not specify whether they targeted over-the-counter or prescribed medications [24,34,40,42,45,46,50–52,59,76–81]. Additionally, five studies addressed both prescription and over-the-counter medications [43,55,68,69,99].

Overall medication errors

This review categorises medication errors results based on the specific stage in the medication use process [18] (Table 3) and according to the geographical location of the included studies [32].

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Stages in the Medication Use Process targeted by Different Studies*.

https://doi.org/10.1371/journal.pone.0322392.t003

In the Europe and Central Asia region, ten studies examined medication errors across multiple stages in the medication use process [38,43,45–47,49,50,56,61,62]. Prescribing errors were the exclusive focus of eight studies [39,41,42,44,48,53,55,57] while seven studies specifically examined dispensing errors [24,40,51,58–60,28]. Additionally, two studies addressed administration errors [52,54].

In North America, three studies assessed medication errors in several stages of the medication use process [77,81,88] while five studies mainly targeted prescribing errors [89,91,93–95], with eight studies examining specifically dispensing errors [80,82–84,86,87,90,92]. Only one study conducted in the US exclusively targeted transcribing errors [85]. In two studies, the specific stage at which the medication error occurs is not mentioned [78,79].

In the Middle East and North Africa (MENA) region, prescribing errors were the exclusive focus of four studies [66,67,70,73] while six studies concentrated solely on dispensing errors [68,69,71,72,74,75].

The included studies from South Asia, Latin America (n = 6) [96–101] and the Caribbean regions (n = 3) [63–65] assessed prescribing errors.

In the East Asia and Pacific region, two studies focused on multiple stages within the medication use process [34,36], one study concentrated on prescribing errors [37] and one study targeted dispensing errors [35].

Only two studies in all regions assessed monitoring errors: one study conducted in the Netherlands [47] and another in Canada [81].

Prescribing errors

This section synthesises data on prescribing errors from studies that focused exclusively on prescribed errors, as well as studies that examined prescribing errors alongside other types of errors. Variations in the rate of prescribing errors were observed across different studies and geographic regions (Table 4). These variations stem from discrepancies in defining error rates and utilising different denominators. Denominators encompassed various metrics including the number of patient visits [37], total number of prescriptions [38,41,44,48,49,53,56,63–66,70,73,89,93,95–97,99–102,104] and total number of items [42,55,57,98]. Within the Europe and Central Asia region, reported error rates ranged from 0.062% in a UK study [55] to 32.84% in a Polish study [53]. The majority of studies (n = 12) addressed both commission and omission errors [38,39,41,42,44,48–50,53,55–57], with commission errors being the most frequently reported (n = 6) [38,39,44,55–57]. The most common commission error was incorrect medication (n = 6) [38,39,44,56,57,62]. Only four studies provided insights into the severity of errors [42,44,55,56], with three studies [42,44,56] reporting errors as ‘non-serious’, while one study highlighted ‘serious’ errors [55].

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Prescribing Errors*.

https://doi.org/10.1371/journal.pone.0322392.t004

In North America, prescribing error rates were reported in three studies [89,93,95] ranging from 4% [95] to 63% [93]. Three studies in this region addressed both commission and omission errors [89,91,95], with omission errors prevailing in two studies [89,95]. Additionally, three studies reported the severity of medication errors [81,89,91], with 50% or more of errors causing harm.

In the MENA region, none of the studies provided data on prescribing error rates. Most studies in this region (n = 3) addressed both commission and omission errors [66,67,70], with omission errors being predominant in two studies [66,67]. Only one study assessed the severity of errors, with 9% of errors reaching patients judged to be likely to have caused harm or resulted in a subtherapeutic effect [66].

Within South Asia, a single study, conducted in India, reported a prescribing error rate of 6.09% (95% CI: 5.78‐6.41) [98]. Most studies in this region (n = 5) solely addressed omission errors [96,97,99–101]; the types of omission errors were variable between the studies. None of these studies documented the severity of errors.

In the East Asia and Pacific region, one study, conducted in Taiwan, reported a prescribing error rate of 18.3% [37]. All studies in this region addressed both commission and omission errors [34,36,37] with commission errors being the most common. Additionally, one study in the Republic of Korea assessed error severity, reporting that the majority of errors (86.4%) were near-miss errors [36].

In Latin America and the Caribbean, all studies focused on omission errors [63–65] with missing mode of administration being the most common error. None of the studies in this region evaluated error severity.

Dispensing errors

This section examines the data on dispensing errors from studies that focused mainly on dispensing errors and studies that assessed both dispensing errors and other types of medication errors. Dispensing errors exhibited considerable variation across studies conducted within or outside the same geographical regions (Table 5). This variation stems from the use of different definitions to categorise a dispensing error and the adoption of different denominators. Denominators employed in the studies included the total number of prescriptions [38,49,56,74,75,82–84,87,92,104], the number of prescribed items [28,58–60], total number of dispensed medications [69,71,72], transactions [86] and total number of patients [35,88]. In 18 studies, the denominator was unspecified [24,40,43,45–47,50,51,61,62,77,81,85,90,80,68,34,36].

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Dispensing Errors*.

https://doi.org/10.1371/journal.pone.0322392.t005

Within the Europe and Central Asia region, the dispensing error rate ranged from 7.1 undocumented dispensing errors per 100,000 dispensed prescriptions in a Finnish study [40] to 3.3% in a study conducted in the UK [28]. Most studies (n = 9) addressed both content and labelling errors [38,50,56,28,58–62] with content errors being the predominant type across all studies. Conversely, five studies focused solely on content errors [24,43,47,49,51]. Among content errors, incorrect dose or strength was the most frequently observed error in six studies [24,28,38,43,60,61], followed by incorrect drug or form in five studies [49,56,58,59,62]. On the other hand, incorrect directions constituted the most common labelling errors [58,59,61,28]. Four studies documented the severity of medication errors [24,38,45,28], with over 50% of errors causing either ‘no harm’ or ‘minor harm’ [24,45,28].

In North America, the range of dispensing error rates varied significantly, from 0.075% [84] to 24% [82]. Among the six studies that evaluated both content and labelling errors [77,82–84,90,80], the latter emerged as the most prevalent issue in three studies [82–84]. Conversely, three studies exclusively focused on content errors [87,88,92]. The most frequently observed content errors included incorrect medication [77,88,90,80] and incorrect quantity [82–84]. Wrong labelling instructions were consistently reported as the most common labelling error for four studies [77,83,84,80]. Additionally, four studies provided insights into the severity of errors using different definitions [81,87,88,80]. For instance, in one study the degree of harm was related to the patient outcome using the following categories: no error, no harm, mild harm, moderate harm, severe harm and death [81]. These studies revealed that the percentage of errors resulting in harm or ADEs varied from 13.6% [87] to 100% [88].

Within the MENA region, the range of dispensing error rates varied from 0.8% in two Yemen-based studies [74,75] to 36.7% in a study conducted in Iran [68]. Among the conducted studies, three exclusively concentrated on content errors [71,74,75], while two addressed both content and labelling errors [69,72]; content errors consistently emerged as the predominant type across all studies [69,71,72,74,75]. Within these studies, incorrect quantity was identified as the most common content error in two studies [69,72], while wrong dosage form prevailed in another two studies [74,75]. Notably, no information regarding the nature of labelling errors was available for the two studies that examined both labelling and content errors [69,72]. Furthermore, two studies provided insights into the severity of dispensing errors, indicating that over 50% of errors were classified as ‘moderate’ to ‘severe’ [69,72].

Only one of the three studies conducted in the East Asia and Pacific region provided a specific figure for the dispensing error rate of 85.4% [35]. Two studies focused on both content and labelling errors, with content errors being the most frequent [34,36]. In the Australian study, the most prevalent content error was identified as incorrect concentration/strength [34], whereas the Korean study highlighted dosing errors as the primary error [36]. No data concerning labelling errors was available for these studies. Only one study targeted labelling errors with illegible medication details being the most common [35]. Furthermore, only one of these studies evaluated the severity of errors, indicating that 3.2% of errors were categorized as ‘mild’ to ‘moderate’ [36].

Administration errors

The data derived from studies that specifically focused on administration errors as well as studies that targeted administration alongside other types of errors were synthesised in this section. Administration errors are summarised in Table 6. Overall, only one study, conducted in Canada, provided data on the administration error rate, revealing that an error occurred in 78% of participants [76], with the incorrect use of inhalers identified as the most prevalent administration error.[76]

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Administration Errors*.

https://doi.org/10.1371/journal.pone.0322392.t006

Within the Europe and Central Asia region, an Irish study reported that ‘doubling the dose’ was the most frequent administration error [43]. Similarly, studies from Spain [52] and Portugal [54], reported the incorrect usage of inhalers as primary administration errors. In the Netherlands, a study found administration of a medication to an incorrect patient to be the most frequent error [47].

In the East Asia and Pacific region, an Australian study comprehensively examined all administration errors [34], with not following prescribed instructions identified as the most common issue[34]In the Republic of Korea, a study targeted errors related to the right patient, right drug, right dose, and right route, with dosing administration errors being the most common [36]. Notably, only this Korean study reported the seriousness of medication errors, revealing that 0.8% of errors resulted in ‘none to mild harm’ [36].

Risk of bias

The risk of bias assessment in included studies are shown in Table 7.

Download:

• PNG
  larger image
• TIFF
  original image

Table 7. Critical Appraisal of Included Studies [22].

https://doi.org/10.1371/journal.pone.0322392.t007

The majority of studies (n = 36) had a low risk of selection bias

[24,35,36,38,41,43,45–49,51,54,55,59,62,63,65,66,69,72,73,77,81,84,86,89,90,92–95,98,100,102,104],

while 24 studies were deemed to have a high risk of selection bias [39,42,50,52,60,61,28,76,78,82,85,87,88,91,80,67,70,71,99,101,34,37,64,103].

Most studies (n = 48) were identified to have a high risk of identification bias [24,38,40–48,] [28,34,36,50–62,65,66,70,71,74–78,81,85,90–93,95,97–99,103,104] whereas 18 studies were classified as having low risk [49,82–84,87–89,94,67–69,72,73,96,101,37,35,64].

Similarly, the majority of studies (n = 48) were categorised as having a high risk of error categorisation bias [24,40,41,43–46,48, 28,36,48,50–62,65,67,68,70,71,73–78,81,85,88,90–93,95,97–99,101,103,104] while 21 studies were deemed to have a low risk [34,35,37–39,42,47,49,64,69,72,80,82,83,86,87,89,94,96,100,102]. Concerning conflict of interest bias, most studies were classified as having a low risk for bias (n = 58) [24,34–42,45–50,53,55,56,58–60,63–65,67–77,79–84,86–92,94,95,97–102,104] with only six studies considered at high risk [43,54,62,85,66,96].

Future aspects

This review provides valuable worldwide insights on medication safety in community pharmacies that can help inform international policy initiatives focused on monitoring, decreasing and preventing medication errors. Moreover, the findings can assist in making informed decisions about where to allocate funding for medication safety initiatives, aimed at alleviating the burden caused by medication errors in different regions and for improving the patient safety culture in community pharmacy settings internationally. One of these initiatives could include the implementation of electronic systems as many reviews have highlighted their effectiveness in decreasing errors within hospital settings [108] and they may also prove effective in community pharmacy settings. Additional strategies encompass setting efficient systems among pharmacy personnel based on teamwork, communication and the no-blame culture [109]. Other strategies include the development of prescribing charts and guidelines [108] alongside process related interventions such as the adoption of the World Health Organisation medication safety guide for look-alike, sound-alike medicines [110]. This review also underscores the role of community pharmacists in adopting patient and medication safety measures. Community pharmacists should be encouraged to implement different strategies and systems to minimise medication errors, including but not limited to enhancing their vigilance in checking prescriptions to mitigate prescription errors, using technology, adopting operational flowcharts or tools, double verifying medications before dispensing, and improving public awareness of the importance of medication safety and reporting medication errors. In fact, there is strong evidence illustrating the impact of pharmacists on reducing medication errors in different healthcare settings. For instance, Gillani et al., highlighted the important role pharmacists play in preventing and in raising awareness about medication errors and in implementing appropriate reporting policies [111]. Moreover, in a systematic review examining the impact of pharmacist interventions on medication errors in hospitalised paediatric patients, Naseralallah et al., concluded that pharmacist involvement can lead to considerable reductions in the overall rate of medication errors [112]. Moreover, as recommended by the IOM, the International Pharmaceutical Federation (FIP) and the WHO [113], all pharmacists and undergraduate pharmacy students should receive comprehensive education and training in pharmacotherapy, and medication safety so that they would have the required knowledge, and skills to detect medication errors and be enabled to act accordingly to resolve issues. [114]Involving pharmacists who possess expertise in medication safety in this setting could also be considered, mirroring the practice observed in hospitals. Educational programs are also required for medical and non-medical prescribers in the outpatient sector to improve their prescribing competency with attention given to the different types of prescribing errors encountered in each region whether omission or commission errors [115].

The scarcity of research on medication errors in community pharmacies in the African and South-East Asian regions highlighted in this review necessitates an urgent attention by policy makers, healthcare professionals and researchers. In fact, according to the 2021 report by the FIP on community pharmacies, the pharmacist-to-pharmacy ratio in these regions is less than one, suggesting that certain pharmacies operate without a pharmacist [116]. Coupled with one of the highest rates of preventable medication harm in these regions, as indicated by the 2024 WHO report on the global burden of preventable medication harm [117], there are growing concerns about medication safety in community pharmacies within these areas. Furthermore, due to the scarcity of comprehensive international data concerning administration, transcribing and monitoring errors, as well as dispensing and prescribing errors in regions where these assessments have been lacking, there is a pressing need to conduct more well-designed studies in these areas.

This review showed that most studies targeted prescription medications with very few targeting over-the-counter or non-prescription medications. Non-prescription medications are accessible directly from pharmacies or other outlets without the need for a prescription. While promoting self-care, the accessibility of non-prescription medications has nurtured a common belief by the general public in relation to the safety of these medications often leading to insufficient awareness regarding their risks of misuse, dependency, and harm. Hence the gap in evidence on over-the-counter medication related errors also needs to be addressed in future studies [118,119].

The results of this review indicate a lack of high-quality research on medication errors in community pharmacies. Therefore, future studies should employ standardised error reporting methods that adhere to universal definitions or taxonomy of medication errors and classifications for error types and severity. Additionally, they should utilise specific denominators to enable the assessment of error rates across studies and should have a thorough analysis of the severity of these errors by assessing the potential harm on patients. Tools that could be used for this purpose include the NCC MERP Index for Categorising Medication Errors, which classifies harm into eight categories [120]. Moreover, the review indicated that policy guidelines, tools or frameworks should be designed for reporting and calculating community pharmacy error rates so that comprehensive data would be accurately and consistently captured. This is necessary for providing a more precise depiction of the safety levels associated with medication errors in community pharmacies globally and will help in identifying the causes of errors and in designing system level interventions for preventing and alleviating errors.